# Claude Memories Full Content Canonical site: https://claudememories.com Last source check: 2026-09-20 Independent educational publication. Not affiliated with Anthropic. ## Signature experience ### The Chips URL: https://claudememories.com/chips/ Specimen: claudememories-chips-001 The memory you use in chat is the same memory Claude Cowork uses. It lives as short topic files under Topics in Memory settings, and you can read, edit, or delete any of them. On Free, Pro, and Max it is on by default across web, desktop, and mobile. Sensitive topics stay off until you opt in. On Team and Enterprise, admins gate availability and it stays off until you enable it. Provenance: Original work by Claude Memories. Every topic name and file body is invented for teaching. No real private memory, medical record, or organization setting appears here. Labeled synthetic topics on the desk: - preferred citation style (preferred-citation-style.md, preference): Use Chicago notes-bibliography for field-guide pages. Short primary-source quotes, then original guidance. Never invent a page number. - project nickname (project-nickname.md, project): Internal nickname for the memory-map rebuild: "the garden desk." Not a client name and not a repository path. - meeting summary format (meeting-summary-format.md, preference): Summaries open with decisions, then owners, then open questions. No recap of the small talk. - activation metric definition (activation-metric.md, work): Activation means the first successful Cowork task in a calendar week, counted in Pacific Time, Monday as week start. - scheduling timezone (scheduling-timezone.md, work): Schedule in Pacific Time. If a time is ambiguous, ask rather than converting silently. - dietary notes (dietary-notes.md, sensitive): Prefers recipes without gluten when suggesting meal prep. Synthetic specimen — not a real medical record. - org writing voice (org-writing-voice.md, managed): On managed plans, shared voice notes wait on two gates: an admin making memory available, then the user turning it on. This specimen is a house style, not an employee file. How to use the real Settings > Memory list: - Open Memory settings: Go to Settings > Memory. That is the control surface for turning memory on, browsing topics, and choosing whether sensitive topics are included. - Read the topic list: Everything Claude remembers appears as short files under Topics. Skim them the way you would skim a small config directory. - Edit a wrong fact in place: Correct the topic file rather than arguing in the next chat. A fix pays off everywhere, including Cowork cloud tasks. - Delete finished work: Remove topics tied to completed projects instead of resetting all memory. Reset is still available, and it is destructive. - Leave sensitive topics off unless you opt in: Health, beliefs, politics, and similar categories stay out of memory until you enable them. The opt-in is not retroactive, and some categories never save. - On Team or Enterprise, check both gates: Confirm an admin has made memory available, then enable it for yourself. It stays off until you do. FAQ: - Are these chips my real Claude memory? No. They are a labeled synthetic specimen authored by Claude Memories. Every topic name and file body is invented. The page does not connect to a Claude account, does not read Settings > Memory, and does not upload anything. - Is the memory in chat the same memory Claude Cowork uses? Yes. Anthropic’s September 2026 announcement states that the memory you use in chat is the same as in Claude Cowork. Context built in conversation is available when Cowork runs a cloud task, and what comes up in Cowork carries back to chat. - Can I delete one thing Claude remembers without resetting everything? Yes. Memories are stored as individual topic files under Topics in Memory settings. Each one can be read, edited, or deleted on its own. Pausing or resetting all memory remains available. - Why does the dietary-notes chip stay off? Sensitive topics such as health, race, ethnicity, religious beliefs, politics, and gender identity are excluded by default. You can turn on “include sensitive topics in memory” in Memory settings. That opt-in saves going forward only; nothing from before is backfilled. Some categories stay excluded either way. - Why is the Team / Enterprise chip gated? On Team and Enterprise, admins control availability for the organization, and memory stays off for individual users until they turn it on. That is two gates, not one. On Free, Pro, and Max, memory is on by default across web, desktop, and mobile. - Is memory everywhere a larger context window? No. It changes what Claude knows at the start of a conversation or Cowork task, not how much it can hold during one and not what a repeated prompt costs. Context-window sizing, compaction, and prompt caching belong on claudecontext.com. - Is Claude Memories affiliated with Anthropic? No. Claude Memories is an independent educational publication and is not affiliated with Anthropic, Claude, or Claude Code. ## Pages ### Claude Memory Everywhere: Chat, Cowork, and Who Controls It URL: https://claudememories.com/memory-everywhere Description: Claude memory now spans chat and Claude Cowork as one editable set of topic files, with sensitive topics off by default and admin gating on Team and Enterprise. Summary: The memory you use in chat is the same memory Claude Cowork uses, so context carries between conversations and cloud tasks. Claude stores it as short topic files under Memory settings, where you can read, edit, or delete any of them. It is on by default for Free, Pro, and Max; sensitive topics are off until you opt in; Team and Enterprise admins gate availability. Quotable callout: One memory - Chat and Claude Cowork now draw on one memory, so what you explain in a conversation is available to a cloud task and what happens in that task carries back. Takeaways: - Chat memory and Claude Cowork memory are the same memory, not two stores that sync. - Memory is a list of short topic files in Settings > Memory that you can read, edit, or delete individually. - Sensitive topics are excluded by default and only save going forward if you opt in. - On by default for Free, Pro, and Max; on Team and Enterprise, admins control availability and it stays off until a user enables it. Sections: - One memory across chat and Cowork: The change Anthropic announced is small to describe and large in practice: Claude Cowork now uses the same memory as chat. Previously, a conversation and a Cowork task were separate contexts, so anything you had explained in chat had to be re-explained before Cowork could act on it. Now, when Cowork runs a task in the cloud, what Claude remembers from your chats is there, and what comes up during the task carries back into chat. The practical effect is that briefing Claude becomes a one-time cost rather than a per-surface cost. If you explain in chat how your team defines a metric, decks and documents that Cowork builds afterwards use that definition. If you brainstorm the shape of an event in chat, Cowork already knows the headcount and the city when it builds the logistics doc. This also means the memory audit habit matters more than it did. One store feeding two surfaces is convenient, but a wrong or stale topic now reaches further. - Memory updates as you chat: Claude now adds topics to memory while you are chatting rather than summarizing a conversation once it ends. Mention that a deadline moved, and the next conversation already knows it without a "remember this" instruction. That is a meaningful shift from the older model, where memory was a synthesis assembled from chat history after the fact. It shortens the delay between telling Claude something and Claude acting on it, and it makes the stored unit a topic rather than a rolling summary of everything. You can still pause memory or reset it at any time. - See, edit, and delete topic by topic: Everything Claude remembers appears as a list of files under Topics in Memory settings. Each file is short, and each one can be read, edited, or deleted on its own. You no longer have to choose between living with a wrong memory and resetting the whole thing. Anthropic's framing of why this matters is worth keeping: a fix pays off everywhere. Correct your company's old name in one topic file and every conversation from then on gets it right, across both chat and Cowork. Use this the way you would use any small config directory. Skim it periodically, delete topics tied to finished work, and correct facts in place rather than arguing with Claude in conversation. - Sensitive topics are off by default: By default, Claude does not store topics related to personal or sensitive subject matter. Anthropic's list includes health, race, ethnicity, religious beliefs, politics, gender identity, and similar areas. The setting is opt-in, not opt-out. If those are the topics you are tired of re-explaining, you can turn on include sensitive topics in memory in Memory settings; Anthropic's example is Claude remembering a gluten allergy when suggesting recipes. With the setting on, you get a notice each time Claude saves something in one of these categories, and the setting can be turned off again at any time. Two limits are worth knowing before you enable it. First, it is not retroactive: Claude saves sensitive topics going forward only, and nothing from before you turned it on is backfilled. Second, some categories stay excluded even with the setting on, including sensitive identification numbers such as Social Security or government ID numbers, criminal history, immigration status, and anything that violates Anthropic's Acceptable Use Policy. Claude tells you when it cannot store something in those categories. - Who has it, and who turns it on: On Free, Pro, and Max , memory is on by default across web, desktop, and mobile. Saving sensitive topics remains off by default on those plans. On iOS and Android, you need the latest app version to get the current behavior. On Team and Enterprise , control moves up a level: admins decide availability for the organization, and memory is off for individual users until they turn it on themselves. That is two gates, not one. If you are on a managed plan and see no memory topics, check whether your admin has enabled it before assuming a bug. The control surface for everyone is Settings > Memory , which is where you turn memory on, browse the topic list, and choose whether sensitive topics are included. - What this is not: Memory everywhere is not a larger context window and not a caching feature. It changes what Claude knows at the start of a conversation or task, not how much it can hold during one or what a repeated prompt costs. For context-window sizing, compaction, and prompt caching, use the sibling resource Claude Context . It is also distinct from Claude Code memory. CLAUDE.md and auto memory are file-based context that lives in your repository and on your machine, written by you and by Claude Code. They are not part of the chat and Cowork topic list, and the September 2026 announcement does not change them. If you are mapping all of these against each other, start with the memory map . FAQ: - Does Claude Cowork have its own separate memory? No. Anthropic says the memory you use in chat is the same one Cowork uses. What Claude remembers from your chats is available when Cowork runs a task, and what comes up in Cowork carries back to chat. - Can I delete one thing Claude remembers without resetting everything? Yes. Memories are stored as individual topic files under Topics in Memory settings, and each one can be read, edited, or deleted on its own. Pausing or resetting all memory is still available. - Does Claude remember health or political information? Not by default. Sensitive categories such as health, race, ethnicity, religious beliefs, politics, and gender identity are excluded unless you turn on "include sensitive topics in memory". Even then, Claude does not store sensitive ID numbers, criminal history, immigration status, or anything violating the Acceptable Use Policy. - Does turning on sensitive topics apply to past conversations? No. Claude saves sensitive topics going forward only. Anything from before you turned the setting on is not saved retroactively, and you can turn the setting off again at any time. - Is memory on for Team and Enterprise users? Not automatically. Admins control availability for the organization, and memory is off for individual users until they turn it on. On Free, Pro, and Max, memory is on by default across web, desktop, and mobile. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - Claude release notes (Claude Help Center): https://support.claude.com/en/articles/12138966-release-notes ### What Is Claude Memory? URL: https://claudememories.com/what-is-claude-memory Description: A plain-language map of Claude app memory, Claude Code memory, project memory, and developer-platform memory. Summary: Claude memory is not one feature. It is a set of persistence surfaces: the topic files shared by chat and Claude Cowork, project memory, Claude Code instruction files, Claude Code auto memory, and platform memory tools. Quotable callout: 5 surfaces - Treat Claude memory as five related surfaces, not one global brain: chat and Cowork topics, project memory, CLAUDE.md, auto memory, and platform memory tools. Takeaways: - Chat and Claude Cowork share one memory, stored as editable topic files in Settings > Memory. - Project memory is scoped separately from non-project memory. - Claude Code starts fresh each session, then loads persistent context from files and auto memory. - Developer-platform memory is a tool pattern controlled by the application builder. Sections: - The short version: Claude can remember across sessions, but the mechanism depends on where you are using Claude. In the Claude app, memory is a set of short topic files built as you chat, and since September 2026 the same memory is used by Claude Cowork. In Projects, memory is separated by project. In Claude Code, memory usually means persistent instruction files and auto-written Markdown notes. On the developer platform, memory can mean an agent tool that stores files outside the context window. The practical mistake is asking "does Claude remember?" without asking "which Claude surface, which scope, and who controls the storage?" A useful memory setup begins by naming the surface first. For the current chat and Cowork behavior, see memory everywhere . - The five memory surfaces: Chat and Cowork memory is the user-facing topic list Claude builds as you chat, shared between conversations and Claude Cowork tasks. Use it for personal work context, stable preferences, and recurring collaboration details. Project memory keeps a project's summary separate from non-project chats. Use it when the memory should not leak into unrelated work. CLAUDE.md is written by a human and loaded as Claude Code context. Use it for rules, conventions, architecture notes, and commands that should be present every session. Auto memory is written by Claude Code itself as plain Markdown. Use it for learned preferences and project facts Claude discovers while working. Platform memory tools are application-level storage patterns. Developers decide where the data lives and how Claude reads or writes it. - Memory is context, not enforcement: Claude Code documentation is explicit that memory files are context rather than hard configuration. That matters. A memory can influence behavior, but if a workflow must run at a fixed time, such as before a commit or before a tool call, use a hook, policy setting, or automation rather than a reminder in a memory file. This distinction explains many "Claude ignored my memory" reports. Persistent context improves odds; it does not make a soft instruction into a guardrail. - What to read next: If you use chat, Cowork, or the mobile app, start with memory everywhere for what changed in September 2026, then Claude app memory for the day-to-day controls. If you use the terminal, read Claude Code memory and then the deeper guides to CLAUDE.md and auto memory . For context-window sizing, compaction, and prompt caching, use the sibling resource Claude Context . For broader Claude Code engineering practice, use Claude Ships Code . FAQ: - Does Claude remember everything I say? No. Chat memory is stored as short topic files you can edit or delete, sensitive topics are excluded by default, and incognito chats are excluded from memory. Claude Code memory is file-based context and auto memory, not a full transcript. - Is Claude memory the same as context window? No. Memory is persistent context that can be loaded or retrieved across sessions. The context window is the active working space for a single conversation or agent run. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - Managing context on the Claude Developer Platform (Claude by Anthropic): https://claude.com/blog/context-management ### Claude App Memory: Settings, Topics, and Controls URL: https://claudememories.com/claude-ai-memory Description: How Claude app memory works across web, desktop, and mobile, including memory topic files, Settings > Memory controls, incognito chats, and reset behavior. Summary: Claude app memory is a controllable set of short topic files Claude writes as you chat, shared with Claude Cowork. It is useful for stable preferences and work context, but it should be audited before sensitive or unrelated work. Quotable callout: Topics, live - Claude now adds topics to memory as you chat instead of summarizing conversations after they end, and each topic is a file you can edit or delete. Takeaways: - Memory is available in Claude web, Desktop, and Mobile, and the same memory is used by Claude Cowork. - Memory is a list of short topic files, each of which can be read, edited, or deleted individually. - Users turn memory on and control what is remembered from Settings > Memory. - Incognito chats are not saved to Claude memory or chat history. Sections: - What the app remembers: Claude app memory is built from topics, not transcripts. As you chat, Claude writes short files covering things worth carrying forward: your role, the projects you are working on, professional context, communication and technical preferences, coding style, and ongoing work. Those categories are work-oriented, and they are not a promise that every fact will be retained or used. The useful mental model is not "Claude has a database of every detail." It is "Claude keeps a small set of notes it can consult in future conversations, and you can read all of them." Since the September 2026 update, the same notes are used by Claude Cowork . - Where to control it: Memory controls live in Settings > Memory . From there you turn memory on, browse the topic list, edit or delete individual topics, choose whether sensitive topics are included, and pause or reset memory. Reset is the destructive option: it permanently deletes memories, including project memories. You can also tell Claude in chat what to remember or what to change. For durable preferences, make the instruction concrete: "remember that I prefer concise meeting summaries with decisions first" is better than "remember my style." Because topics are editable files, correcting a fact once fixes it everywhere rather than requiring a correction in each new conversation. - Incognito, sensitive topics, and deletion: Incognito chats are the clean-room option. When enabled, Claude will not save the chat to memory or chat history. The Help Center also describes deleted conversations being removed from memory. Sensitive categories are excluded by default. Claude does not store topics about health, race, ethnicity, religious beliefs, politics, gender identity, or similar areas unless you turn on the sensitive-topics setting, and some categories stay excluded either way. The memory everywhere guide covers the exact boundaries. For sensitive work, use both product controls and ordinary operational hygiene: confirm memory state, skim the topic list, use incognito where appropriate, and do not paste secrets or regulated data unless your plan and policies explicitly allow it. - What not to store: Memory should hold stable collaboration context, not passwords, access tokens, private customer data, or one-off facts that will expire quickly. If a fact has a short shelf life, store it in the live project artifact instead: an issue, task tracker, runbook, or source-controlled doc. When memory and the source of truth disagree, the source of truth should win. Memory is a convenience layer, not a records system. FAQ: - Where do I edit Claude app memory? Use Settings > Memory. Each memory is a short topic file you can read, edit, or delete. You can also tell Claude in chat what to remember or change. - Does incognito chat affect memory? Yes. Claude Help says incognito chats are not saved to memory or chat history. - Is app memory the same memory Claude Cowork uses? Yes. Since September 2026 the memory you use in chat is the same one Cowork uses, so context moves in both directions between conversations and cloud tasks. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - Bringing memory to teams (Claude by Anthropic): https://claude.com/blog/memory - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects ### Claude Project Memory: Boundaries, Instructions, and RAG URL: https://claudememories.com/project-memory Description: How Claude project memory summaries differ from profile instructions, project instructions, and project knowledge retrieval. Summary: A Claude Project gives memory a boundary. It can have its own memory summary, its own instructions, and uploaded knowledge, but those are different mechanisms. Quotable callout: 1 project, 1 summary - For Team and Enterprise memory workflows, Claude Help says each individual project has a separate memory summary. Takeaways: - Project memory summaries are separate from non-project chat memory. - Moving a chat into or out of a project changes which summary it belongs to. - Project instructions guide behavior inside a project; they are not the same as memory summary. - Project RAG retrieves uploaded knowledge when project content grows large; it is not the same as memory. Sections: - Three project layers: Claude Projects combine three ideas that are often described as one thing. Project memory is the summary Claude maintains for that project. Project instructions are guidance you write for chats inside that project. Project knowledge is uploaded material Claude can use, and in larger projects it may be retrieved with RAG. Keeping those layers separate makes project behavior easier to debug. If Claude gets a fact wrong, check project knowledge. If Claude uses the wrong tone, check project instructions. If Claude carries stale assumptions across chats, inspect memory. - Moving chats changes the memory boundary: The Help Center describes moving chats into and out of Projects as a way to manage what is included in memory summaries. If a conversation happened in the wrong place, move it out so it no longer belongs to that project-specific summary. This is especially important for agencies, consulting teams, and anyone switching between clients. Project boundaries are only useful when chats are filed in the right project. - Project instructions vs project memory: Use project instructions for stable rules: audience, role, workflow, terminology, formatting expectations, and constraints. Use memory for learned continuity: what the project has been doing lately, what decisions have been made in chats, and what collaboration preferences are emerging. If something must be visible before the first chat in a project, write it as an instruction. If it emerges over time, memory can carry it, but you should still promote durable decisions into a project doc. - RAG is not memory: Claude Help describes project RAG as a way to retrieve relevant information from uploaded documents instead of loading all project content into memory at once. That means RAG is a retrieval strategy for project knowledge, not the same thing as the memory summary. Use clear filenames and organized uploads for knowledge. Use project instructions for behavior. Use memory for continuity. Blending all three into one "memory" bucket is how projects become hard to reason about. - Projects after the September 2026 memory update: Anthropic's memory everywhere announcement changed how chat memory is stored and where it applies: memory is now a set of topic files in Settings > Memory, built as you chat, and shared with Claude Cowork. The announcement does not restate project-scoped memory behavior. The honest reading is that the Help Center's description of per-project memory summaries remains the current reference for Projects, and the topic list is the current reference for chat. Before you rely on a project boundary for anything sensitive, open Settings > Memory and read the topics that actually exist on your account rather than assuming where the line falls. FAQ: - Should I put project rules in memory or project instructions? Put stable rules in project instructions. Let memory capture continuity, then promote important decisions into durable docs. - Does uploading documents create memory? No. Uploaded project knowledge can be retrieved, including through RAG when needed, but it is distinct from memory summaries. Sources: - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - Understanding Claude's personalization features (Claude Help Center): https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features - Retrieval augmented generation (RAG) for projects (Claude Help Center): https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projects - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it ### Claude Code Memory: The Working Map URL: https://claudememories.com/claude-code-memory Description: The practical map of Claude Code memory: CLAUDE.md, CLAUDE.local.md, .claude/rules, auto memory, and /memory. Summary: Claude Code memory is a startup-context system plus a local Markdown memory directory. The core split is human-written instructions versus Claude-written auto memory. Quotable callout: 2 mechanisms - Claude Code carries knowledge across sessions through human-written CLAUDE.md files and Claude-written auto memory. Takeaways: - Every Claude Code session starts with a fresh context window. - Human-written instructions live in CLAUDE.md, CLAUDE.local.md, and .claude/rules. - Claude-written notes live in auto memory under the user home directory. - /memory is the first diagnostic command for loaded files and auto memory. Sections: - The startup model: Claude Code does not simply continue with the same hidden state forever. The docs say each session begins with a fresh context window. Persistent knowledge is then loaded into that new session from memory sources. That means memory quality affects session quality. A precise CLAUDE.md gives Claude a good launch. A bloated, contradictory file gives Claude noisy context. A stale auto memory can keep reintroducing a fixed bug or outdated command. - Human-written memory: The human-written side includes CLAUDE.md , CLAUDE.local.md , and rule files under .claude/rules/ . Use this layer for instructions you intentionally want Claude to see: commands, architecture notes, code style, testing expectations, and repo-specific caveats. Claude Code reads CLAUDE.md files from multiple scopes, including managed, user, project, and local locations. Project files are shareable through version control; local files are for personal preferences and should stay gitignored. - Claude-written memory: Auto memory lets Claude Code write notes for itself. The docs describe notes such as build commands, debugging insights, architecture notes, code style preferences, and workflow habits. This is useful because not every durable learning is obvious before a session starts. Because auto memory is Claude-written, audit it. Ask what it saved, run /memory , and open the folder when behavior starts feeling haunted by old assumptions. - Use /memory first: The /memory command lists loaded CLAUDE.md , CLAUDE.local.md , and rules files. It also lets you toggle auto memory and open the auto memory folder. If Claude is missing a rule, start by proving the file is actually loaded. When a rule is loaded but ignored, rewrite it more concretely, remove conflicts, or move hard requirements into hooks or settings. FAQ: - What is the difference between CLAUDE.md and auto memory? CLAUDE.md is written by you for persistent instructions. Auto memory is written by Claude based on corrections, preferences, and useful project learnings. - Can memory enforce behavior? No. Claude Code docs describe memory as context. Use hooks or settings for enforcement. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### CLAUDE.md Guide: What to Put In It and Where URL: https://claudememories.com/claude-md Description: A practical, source-grounded guide to writing CLAUDE.md files for project instructions, local preferences, imports, and AGENTS.md bridges. Summary: A good CLAUDE.md is short, specific, and useful every session. It should explain how the project works, not become a dumping ground for every task note. Quotable callout: < 200 lines - Anthropic recommends targeting under 200 lines per CLAUDE.md because larger files consume more context and may reduce adherence. Takeaways: - Use project CLAUDE.md for shared repo behavior and commands. - Use CLAUDE.local.md for personal preferences, local URLs, and private test data. - Use .claude/rules for path-specific or modular instructions. - Use imports for organization, but remember imported files still load into context. Sections: - What belongs in CLAUDE.md: Put the facts you would otherwise repeat at the start of every Claude Code session: build commands, test commands, project layout, code style, deployment caveats, known slow tests, and review expectations. The best entries are concrete enough to verify. Instead of "be careful with tests," write "run npm run build before deploy and use focused tests while iterating." Instead of "follow our style," write the exact formatting, naming, or architectural rule that matters. - Choose the narrowest scope: Claude Code supports multiple locations. Use managed policy files for organization-wide requirements, user files for personal defaults, project files for team-shared repo guidance, and local files for machine-specific or private details. Do not put personal tokens, customer data, or local-only URLs into a project file that will be committed. Use CLAUDE.local.md and keep it gitignored. - Imports and AGENTS.md: CLAUDE.md can import other files with @path/to/file . That is useful for organization, but it does not reduce context cost because imports are expanded at launch. If a long procedure only matters sometimes, move it to a skill or path-scoped rule instead. If a repo already uses AGENTS.md , Claude Code docs say Claude reads CLAUDE.md , not AGENTS.md . The bridge is simple: create a CLAUDE.md that imports @AGENTS.md , then add Claude-specific notes below it. - A compact starter shape: # Project Instructions ## Commands - Install: npm install - Build: npm run build - Test: npm test ## Architecture - Pages live in src/pages. - Shared components live in src/components. - Data models live in src/data. ## Working Rules - Keep edits scoped to the requested behavior. - Run npm run build before deploying. - Do not change generated files by hand. This is intentionally small. Add only instructions that improve future sessions. Remove rules that become obvious from the codebase or drift out of date. FAQ: - Should CLAUDE.md include long documentation? Usually no. Link or import only what must load every session. Use path-scoped rules or skills for long, conditional procedures. - Does importing AGENTS.md make Claude Code read it? Yes. Claude Code reads CLAUDE.md, and a CLAUDE.md can import AGENTS.md with @AGENTS.md. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Claude Code Auto Memory: Storage, Limits, and Audits URL: https://claudememories.com/auto-memory Description: How Claude Code auto memory works, where MEMORY.md lives, what loads at session start, and how to audit or clean it. Summary: Auto memory is Claude Code writing local Markdown notes for itself. It is useful precisely because it is automatic, and risky precisely because it can become stale. Quotable callout: 200 lines / 25KB - Claude Code loads the first 200 lines or first 25KB of MEMORY.md at session start, whichever comes first. Takeaways: - Auto memory stores plain Markdown in ~/.claude/projects//memory/. - Auto memory is on by default in current Claude Code releases and can be toggled with /memory or settings. - MEMORY.md is the startup index; topic files are read on demand. - Auto memory is per repository and shared across worktrees. Sections: - What auto memory is: Auto memory lets Claude Code accumulate knowledge without you editing a project instruction file. Current Claude Code docs say auto memory is on by default, and you can toggle it from /memory , project settings, or an environment variable. It can save build commands, debugging insights, architecture notes, code style preferences, and workflow habits when Claude decides they will help later. That makes auto memory a learning layer, not a policy layer. It should complement CLAUDE.md , not replace it. - Where it lives: The docs describe auto memory files as plain Markdown under ~/.claude/projects/<project>/memory/ . The project path is derived from the git repository, so worktrees and subdirectories in the same repo share one auto-memory directory. MEMORY.md acts as the concise entrypoint. Deeper topic files can hold more detail and are read when needed. Because these files are local to your machine and outside the repo by default, they are good for personal learned context. If a learning should help the whole team, promote it into a shared project file or docs page. - The startup limit: Claude Code loads the first 200 lines or first 25KB of MEMORY.md at conversation start, whichever comes first. Content beyond that threshold is not automatically loaded at startup. Topic files are available on demand, but only if Claude needs and reads them. That means MEMORY.md should be an index, not a warehouse. Keep it short, point to topic files, and delete stale notes. - Audit rhythm: Audit auto memory when Claude repeats an old mistake, when a repo changes its build system, after a major migration, before switching clients, and before handing a workstation to someone else. A simple audit loop is: run /memory , open the auto memory folder, skim MEMORY.md , remove stale facts, and promote team-relevant items into CLAUDE.md or project docs. FAQ: - Is auto memory committed to git? No by default. Claude Code auto memory lives under your home directory, not inside the repository. - Can I edit auto memory? Yes. The docs describe auto memory files as plain Markdown that you can edit or delete. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Claude Memory Hygiene: Keep Persistent Context Useful URL: https://claudememories.com/memory-hygiene Description: A maintenance guide for keeping Claude memory concise, current, scoped, and safe across app memory and Claude Code memory. Summary: Memory is only useful while it is more current than your next prompt. Healthy memory is short, scoped, auditable, and easy to override with live source truth. Quotable callout: Promote or prune - Every durable memory should either be promoted into a source of truth, kept because it guides future sessions, or pruned as stale context. Takeaways: - Separate stable rules from emergent learnings. - Prefer project-scoped memory over global memory for client or domain-specific work. - Do not store secrets, credentials, or regulated data in memory. - Schedule memory audits around migrations, team changes, and sensitive work. Sections: - Memory decays: Persistent context feels like a shortcut, but every shortcut can become wrong. A build command changes. A project decision is reversed. A style preference stops applying. The longer a memory lives without review, the more likely it is to be plausible but stale. Use a simple standard: if a memory is a true project rule, promote it into the project source of truth. If it is a personal preference, keep it scoped personally. If it is a temporary fact, delete it after the task ends. Two things raised the stakes in September 2026. Chat memory is now stored as individually editable topic files, which makes correcting a stale fact cheap. And the same memory is used by Claude Cowork , which means a stale topic no longer affects only your next conversation — it affects the cloud tasks Claude runs on your behalf. - Scope by risk: Global memory should contain only things that are safe across nearly every conversation: broad communication preferences, stable role context, and durable tooling preferences. Project memory should contain project-specific continuity. Local Claude Code memory should contain personal machine or repo learnings. Client names, legal matters, unreleased product details, and credentials deserve stricter boundaries. Do not rely on memory to enforce data separation; use the right project, incognito mode, workspace controls, and organizational policy. - Write better memories: Good memories are concrete, dated when needed, and easy to test. "The app uses pnpm" is useful. "The app probably uses the new stack" is dangerous. "As of 2026-07-06, production deploys use Cloudflare Pages via wrangler" is better when the fact may drift. For Claude Code, short bullets beat essays. For Claude app memory, tell Claude what to remember in direct language and verify the summary afterward. - Audit checklist: Read the topic list in Settings > Memory before sensitive standalone chats, and delete topics tied to finished work. Correct a wrong fact by editing its topic file rather than restating it in conversation. Move wrongly filed chats out of Projects. Run /memory when Claude Code behavior surprises you. Prune stale MEMORY.md facts after migrations. Promote team-relevant local learnings into shared docs. Use hooks or settings for requirements that must be enforced. FAQ: - How often should I audit Claude memory? Audit after major project changes, before sensitive work, and whenever Claude repeats stale assumptions. - Should memory store secrets? No. Store credentials in proper secret managers, not Claude memory or instruction files. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Claude Memory Setup Checklist URL: https://claudememories.com/claude-memory-setup Description: A practical setup checklist for Claude memory across app settings, Projects, CLAUDE.md, local memory, and auto-memory files. Summary: Set up Claude memory by choosing the narrowest surface that fits the job: app memory for personal continuity, Projects for scoped chat memory, CLAUDE.md for shared repo context, and auto memory for local learned notes. Quotable callout: Scope first - Choose the memory boundary before writing the memory. Most failures come from putting a local, project, or team fact in the wrong surface. Takeaways: - Start with the memory surface: app, project, shared repo, local repo, or auto memory. - Use Settings > Memory for Claude app and Cowork memory, and /memory for Claude Code memory. - Keep shared CLAUDE.md files short, concrete, and safe to commit. - Use tools to generate starter files, design a memory scaffold, and estimate context cost. Sections: - The setup sequence: Name the scope. Decide whether the memory belongs globally, inside a Claude Project, inside a shared repository, or only on your machine. Write the durable rule once. Put shared repo facts in CLAUDE.md , personal repo facts in CLAUDE.local.md , and path-specific guidance in .claude/rules/ . Let learned local facts stay local. Use auto memory for recurring local learnings, then audit it with /memory . Promote what becomes official. If an auto memory becomes a team rule, move it into a source-controlled instruction file or docs page. - App and Project memory: Use Claude app memory for stable personal work context and collaboration preferences. Open Settings > Memory to turn it on, skim the topic list before sensitive work, and use incognito when a chat should not contribute to memory or chat history. Remember that the same memory is used by Claude Cowork, so anything stored there reaches cloud tasks too. On Free, Pro, and Max, memory is on by default. On Team and Enterprise, an admin has to enable it for the organization first, and it stays off for you until you turn it on. Sensitive topics are off by default on every plan. Use Claude Projects when context should stay separated by client, team, or initiative. Project memory summaries are separate from non-project memory, and moving a chat into or out of a Project changes which summary it belongs to. - Claude Code memory: For Claude Code, begin with a compact CLAUDE.md . Include commands, layout, coding standards, and review expectations that should load every session. Put private local details in CLAUDE.local.md , and use rules for path-specific guidance. Then let auto memory carry learned notes that are useful later but not yet team policy. Because auto memory is plain Markdown, you can read, edit, and delete it when behavior drifts. - Free setup tools: Use the CLAUDE.md Generator to draft a starter file, the Memory File Designer to choose a scaffold, and the Context Budget Estimator to keep always-loaded context under control. The tools are browser-only. They do not send your project details anywhere, and they are best treated as setup aids rather than authority over your real repo. FAQ: - What is the safest first memory to create for Claude Code? Create a short project CLAUDE.md with shared commands, layout, and conventions. Add local-only or private details to CLAUDE.local.md instead. - Should I enable every Claude memory feature? No. Enable and write only the surfaces that match your workflow, privacy needs, and plan controls. - What should I do before a sensitive project? Read the topic list in Settings > Memory, use the right Project or incognito mode, run /memory in Claude Code, and remove stale or overbroad local memory. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Claude Memory Troubleshooting URL: https://claudememories.com/troubleshooting Description: A diagnostic checklist for Claude memory not loading, Claude ignoring CLAUDE.md, stale auto memory, compaction surprises, and project boundary mistakes. Summary: Most Claude memory problems reduce to four questions: was the right memory enabled, was the right file loaded, is the instruction specific, and is another memory contradicting it? Quotable callout: 4 questions - Enabled? Loaded? Specific? Conflicted? Those four questions solve most Claude memory troubleshooting sessions. Takeaways: - Use Settings > Memory for Claude app and Cowork memory checks. - Use /memory for Claude Code loaded-file and auto-memory checks. - Rewrite vague instructions into verifiable commands. - Use hooks or settings for requirements that must always run. Sections: - If Claude app memory seems wrong: Start in Settings > Memory. Confirm memory is enabled, read the topic list, and edit or delete the specific topic that is wrong. Editing one file is almost always better than resetting everything, because reset is permanent and takes project memories with it. If you see no topics at all on a Team or Enterprise plan, the cause may be availability rather than a bug: admins gate memory for the organization, and it stays off for individual users until they turn it on. If a whole category is missing, check whether it falls under sensitive topics, which are excluded by default. If the issue involves a specific Project, check whether the relevant chat belongs in that project or should be moved out. - If Claude Code misses an instruction: Run /memory . Verify the expected CLAUDE.md , CLAUDE.local.md , and rule files are listed. If a file is absent, check where Claude Code was launched, whether the file lives in an ancestor directory, and whether exclusions or scope rules apply. If the file is loaded, check whether another instruction contradicts it. Claude Code loads multiple files into one context; closer or local files may appear later, but contradictions still create unreliable behavior. - If the memory is too vague: Replace soft preferences with observable rules. "Format cleanly" is weak. "Use 2-space indentation in TypeScript files" is stronger. "Always test before deploy" is weak. "Run npm run build before wrangler pages deploy " is stronger. Claude reads structure the way people do. Use short headers, bullets, and commands. Remove archive notes that do not matter every session. - If instructions disappear after /compact: Claude Code docs say project-root CLAUDE.md survives compaction because Claude re-reads it from disk and reinjects it. Nested files in subdirectories may not reappear until Claude reads files in those subdirectories again. If the lost instruction only lived in conversation, it was not durable memory. Add it to the right memory surface before compacting or starting a new session. FAQ: - Why is Claude not following CLAUDE.md? The file may not be loaded, the instruction may be vague, another instruction may conflict, or the requirement may need enforcement through hooks or settings. - Does /compact erase CLAUDE.md? Project-root CLAUDE.md is re-read after /compact. Nested files may reload only when Claude reads files in those subdirectories. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects ### Claude Memory Freshness Log URL: https://claudememories.com/freshness-log Description: A dated update log for Claude memory behavior, source checks, and future changes to Claude app and Claude Code memory. Summary: Claude memory features change. This log records source checks and gives future editors a weekly surface to extend. Quotable callout: Dated claims - Availability, settings names, and Claude Code memory behavior should be dated or rechecked before being treated as current. Takeaways: - The current source check is 2026-09-19, covering the memory-everywhere announcement. - Future edits should update this page before changing memory availability claims. - Release notes are the first stop for product changes; feature docs are the first stop for behavior details. - Pages should prefer dated statements over timeless claims when availability can drift. Sections: - 2026-09-19 memory-everywhere source check: Anthropic published "Claude's memory works everywhere, and you decide what's in it" . The changes that affect this site: Chat memory and Claude Cowork memory are the same memory. Context built in conversation is available when Cowork runs a cloud task, and task context carries back to chat. Memory updates as you chat. Claude adds topics during a conversation rather than summarizing it after it ends. Memory is a list of topic files. Under Topics in Memory settings, each memory can be read, edited, or deleted individually. Sensitive topics are off by default. Health, race, ethnicity, religious beliefs, politics, gender identity, and similar areas are excluded unless the user opts in, and the opt-in is not retroactive. Sensitive ID numbers, criminal history, immigration status, and AUP-violating content stay excluded either way. Availability. On by default for Free, Pro, and Max across web, desktop, and mobile. On Team and Enterprise, admins control organization availability and memory is off for individual users until they turn it on. The settings path is Settings > Memory , replacing the older Settings > Capabilities wording used across this site. Editorial decisions from this pass: added memory everywhere as the flagship guide for the change; rewrote the Claude app memory guide around topic files instead of a single after-the-fact summary; retired the "memory updates within 24 hours" timing claim, which described the superseded summary model; and added an explicit note to the Projects guide that the announcement does not restate project-scoped behavior, so the Help Center remains the reference there. Later the same day, added The Chips , a labeled synthetic topic desk where you can inspect, edit, and delete invented memory files. It is original work by Claude Memories, not a real account. Windows, compaction, and prompt caching remain on claudecontext.com . - 2026-07-06 source check: Initial site build. Checked Claude Code memory docs, Claude app memory Help Center article, Projects Help Center article, personalization Help Center article, Project RAG Help Center article, Anthropic context-management post, and Claude release notes. Editorial decision: state app memory behavior from current support pages, and keep older "Bringing memory to teams" launch framing as product background rather than the latest availability source. - 2026-07-06 Phase 2 tool check: Phase 2 added browser-only tools and rechecked the memory behavior used by those tools. Current Claude Code docs state that auto memory is on by default, requires Claude Code v2.1.59 or later, stores project memory under ~/.claude/projects/<project>/memory/ by default, shares that directory across worktrees in the same git repo, and loads the first 200 lines or 25KB of MEMORY.md at session start. The setup and context-budget tools also rechecked Claude Help Center memory behavior: app memory uses an editable summary, project memory spaces are separate from non-project chats, incognito chats do not contribute to memory or chat history, and memory updates within 24 hours after conversations are created, modified, or deleted. Superseded in part by the 2026-09-19 entry above. The summary model and the 24-hour update timing described here were replaced by topic files written during the conversation. The Claude Code auto-memory findings in this entry still stand. - Watch list: Memory availability by plan and surface, including Team and Enterprise admin controls. Whether project-scoped memory keeps a separate summary now that chat memory is stored as topic files. Sensitive-topic category list and the always-excluded categories. Settings navigation names. Claude Code auto memory rollout status and default behavior. MEMORY.md startup load limits. Project RAG capacity and retrieval behavior. New controls for export, deletion, enterprise retention, and incognito behavior. - Update procedure: When updating the site, change the source note first, then the affected article, then this freshness log. If a fact comes from a non-primary community source, label it as workflow advice rather than official behavior. For weekly updates, scan Claude release notes and the memory-related docs. If nothing changed, record "reviewed, no content change" with the date. FAQ: - Why does this site need a freshness log? Claude memory behavior and availability can change, so dated source checks make claims easier to audit. - What should future editors update first? Update source notes first, then page content, then the freshness log. - What is the most recent change to Claude memory? The September 2026 memory-everywhere update: chat and Claude Cowork share one memory, stored as editable topic files, with sensitive topics off by default and admin gating on Team and Enterprise. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - Claude release notes (Claude Help Center): https://support.claude.com/en/articles/12138966-release-notes - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Claude Memory Sources URL: https://claudememories.com/sources Description: The primary-source bibliography behind ClaudeMemories.com, with notes on what each source is used to support. Summary: The site cites Claude-owned documentation and support pages first. Community workflows can inspire future examples, but they do not define official behavior. Quotable callout: 9 - This build uses Claude or Anthropic-owned sources for behavior, availability, controls, and troubleshooting claims. Takeaways: - Primary sources define product behavior. - The site was last source-checked on 2026-09-20. - Release notes and support articles should be checked before changing availability claims. - Community guides are intentionally not treated as authority for storage or deletion behavior. Sections: - Bibliography: Claude memory changes across product surfaces, so this site treats source freshness as part of the content rather than a hidden editorial chore. Every guide prefers Claude-owned docs, help-center pages, release notes, and Anthropic blog posts. Community posts are useful for workflow inspiration, but they are not used as authority for what Claude stores, loads, deletes, or exposes. When a page gives a recommendation, the recommendation is separated from the official behavior it relies on. Claude by Anthropic: Claude's memory works everywhere, and you decide what's in it . Primary reference for shared chat and Cowork memory, topic-level edit and delete, sensitive-topic defaults, and plan availability. Anthropic Claude Code Docs: How Claude remembers your project . Primary reference for CLAUDE.md, auto memory, /memory, file loading, rules, imports, and troubleshooting. Claude Help Center: Use Claude's chat search and memory to build on previous context . Primary reference for Claude app memory, memory summaries, incognito behavior, settings controls, and data handling. Claude by Anthropic: Bringing memory to teams . Earlier launch context for team memory and project boundaries. Its single-summary framing predates the September 2026 move to topic files. Claude Help Center: How can I create and manage projects? . Primary reference for moving chats into or out of Projects and managing project-specific memory summaries. Claude by Anthropic: Managing context on the Claude Developer Platform . Primary reference for developer-platform memory as a file-based tool managed by the developer. Claude Help Center: Understanding Claude's personalization features . Primary reference for profile instructions, project instructions, and styles. Claude Help Center: Retrieval augmented generation (RAG) for projects . Primary reference for project knowledge RAG and how it differs from memory summaries. Claude Help Center: Claude release notes . Freshness reference for ongoing Claude app and Claude Code product changes. - Citation policy: Each guide includes a source block with accessed dates. Short primary-source quotes are included as signposts, while the substantive guidance is original writing derived from the source behavior. If a source changes, the affected article and freshness log should be updated together. FAQ: - Are community posts used as sources? Not for official behavior. They may inform workflow examples in the future, but storage, loading, deletion, and availability claims come from primary sources. - When were these sources checked? The initial source check was completed on 2026-07-06. The current check, covering the memory-everywhere announcement, was completed on 2026-09-20. Sources: - Claude's memory works everywhere, and you decide what's in it (Claude by Anthropic): https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - Bringing memory to teams (Claude by Anthropic): https://claude.com/blog/memory - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects - Managing context on the Claude Developer Platform (Claude by Anthropic): https://claude.com/blog/context-management - Understanding Claude's personalization features (Claude Help Center): https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features - Retrieval augmented generation (RAG) for projects (Claude Help Center): https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projects - Claude release notes (Claude Help Center): https://support.claude.com/en/articles/12138966-release-notes ## Free Tools ### Claude Memory Tools URL: https://claudememories.com/tools/ Description: Free client-side tools for CLAUDE.md generation, memory scaffold design, and context budget estimation. ### CLAUDE.md Generator URL: https://claudememories.com/tools/claude-md-generator Description: Generate a concise, source-grounded CLAUDE.md template for Claude Code projects, with commands, conventions, memory guidance, and safety notes. Summary: Build a compact CLAUDE.md from project facts you provide. The generated file follows current Claude Code guidance: short, specific, scoped, and easy to audit. Key facts: - No prompt, project name, command, or convention leaves your browser. - Claude Code docs recommend targeting under 200 lines per CLAUDE.md file. - CLAUDE.md is context, not a hard enforcement layer; use hooks or settings for requirements that must always run. Features: - Creates a markdown CLAUDE.md template from project type, stack, commands, conventions, and style rules. - Keeps the result focused on instructions Claude should see every session. - Includes a local-memory boundary note so private machine details stay out of committed files. - Copies or downloads the markdown directly in the browser. How to use: - Describe the project type, stack, commands, and conventions. - Choose whether the file is for a shared repository, local-only workflow, or an AGENTS.md bridge. - Generate the markdown, edit anything project-specific, then download or copy it. - Commit only shared project guidance; keep personal or private notes in CLAUDE.local.md. FAQ: - Is this a complete replacement for reading the Claude Code docs? No. It is a practical starter file based on current docs. Review it against your repo and remove anything that is not useful every session. - Should I paste secrets or private URLs into the generator? No. The tool runs locally in your browser, but a committed CLAUDE.md should not contain credentials, customer data, or private machine details. - Why does the generated file stay short? Claude Code documentation says larger CLAUDE.md files consume context and may reduce adherence, so the generator favors concise, concrete rules. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory ### Memory File Designer URL: https://claudememories.com/tools/memory-designer Description: Design a persistent Claude memory scaffold for CLAUDE.md, CLAUDE.local.md, .claude/rules, and auto-memory notes without sending data to a server. Summary: Plan a memory directory that separates shared instructions, local preferences, project rules, and Claude-written auto memory notes before they become tangled. Key facts: - Claude app memory, Claude Project memory, CLAUDE.md, and auto memory are separate surfaces. - Auto memory files are plain Markdown under ~/.claude/projects//memory/ unless configured otherwise. - Project memory summaries are separate from non-project chat memory in Claude Projects. Features: - Builds a recommended file tree for shared, local, path-scoped, and auto-memory notes. - Generates starter markdown for each selected memory surface. - Exports a scaffold plan and an optional shell script to create files locally. - Flags what should stay out of memory, including secrets and short-lived facts. How to use: - Name the project and select the memory surfaces you actually use. - Choose team-sharing and sensitivity settings. - Generate the scaffold, then review each file before creating it. - Promote team-relevant learnings into shared docs and keep private details local. FAQ: - Does this create files on my machine automatically? No. It generates a scaffold plan and an optional shell script. You decide what to run or copy into your project. - Should every project use all memory surfaces? No. Use the narrowest useful surface. A small solo repo may need only CLAUDE.md; a larger team may benefit from rules and local-only notes. - Can this design Claude app memory too? It can help you plan what belongs in app memory versus project or code memory, but app memory itself is managed in Claude settings and chats. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory - Use Claude's chat search and memory to build on previous context (Claude Help Center): https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context - How can I create and manage projects? (Claude Help Center): https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects ### Context Budget Estimator URL: https://claudememories.com/tools/context-budget Description: Estimate the token cost of CLAUDE.md and memory files, see what loads at startup, and spot oversized persistent context before it slows a session down. Summary: Paste your instruction and memory files to estimate token cost, startup load behavior, and the point where memory should move into path-scoped rules or on-demand topic files. Key facts: - CLAUDE.md files loaded at launch consume context alongside the conversation. - Auto memory loads the first 200 lines or 25KB of MEMORY.md at session start. - Imported CLAUDE.md files help organization, but they still load into context at launch. Features: - Estimates tokens from pasted text using a conservative browser-side approximation. - Separates CLAUDE.md startup context from auto-memory MEMORY.md startup limits. - Shows line count, byte size, estimated tokens, and percentage of a selected context budget. - Suggests when to trim, split, or move content into path-scoped rules or topic files. How to use: - Paste one or more memory files into the estimator. - Choose whether each file is CLAUDE.md, auto-memory MEMORY.md, a rule, or an on-demand topic file. - Review estimated tokens, line count, byte size, and startup load status. - Trim always-loaded context until only durable, high-signal guidance remains. FAQ: - Is the token count exact? No. It is an estimate for planning. Tokenization varies by model and content, so treat the result as a budget signal rather than an exact bill. - What context window size should I choose? Choose the context budget that matches your plan and model when you know it. If you are unsure, use the default as a rough planning baseline. - Why does MEMORY.md show a startup cap? Claude Code docs say auto memory loads only the first 200 lines or 25KB of MEMORY.md at session start, with topic files read on demand. Sources: - How Claude remembers your project (Anthropic Claude Code Docs): https://code.claude.com/docs/en/memory - Managing context on the Claude Developer Platform (Claude by Anthropic): https://claude.com/blog/context-management ## Source Bibliography - Claude's memory works everywhere, and you decide what's in it. Claude by Anthropic. https://claude.com/blog/claudes-memory-works-everywhere-and-you-decide-whats-in-it. Accessed 2026-09-19. Primary reference for shared chat and Cowork memory, topic-level edit and delete, sensitive-topic defaults, and plan availability. - How Claude remembers your project. Anthropic Claude Code Docs. https://code.claude.com/docs/en/memory. Accessed 2026-07-06. Primary reference for CLAUDE.md, auto memory, /memory, file loading, rules, imports, and troubleshooting. - Use Claude's chat search and memory to build on previous context. Claude Help Center. https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context. Accessed 2026-07-06. Primary reference for Claude app memory, memory summaries, incognito behavior, settings controls, and data handling. - Bringing memory to teams. Claude by Anthropic. https://claude.com/blog/memory. Accessed 2026-07-06. Earlier launch context for team memory and project boundaries. Its single-summary framing predates the September 2026 move to topic files. - How can I create and manage projects?. Claude Help Center. https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects. Accessed 2026-07-06. Primary reference for moving chats into or out of Projects and managing project-specific memory summaries. - Managing context on the Claude Developer Platform. Claude by Anthropic. https://claude.com/blog/context-management. Accessed 2026-07-06. Primary reference for developer-platform memory as a file-based tool managed by the developer. - Understanding Claude's personalization features. Claude Help Center. https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features. Accessed 2026-07-06. Primary reference for profile instructions, project instructions, and styles. - Retrieval augmented generation (RAG) for projects. Claude Help Center. https://support.claude.com/en/articles/11473015-retrieval-augmented-generation-rag-for-projects. Accessed 2026-07-06. Primary reference for project knowledge RAG and how it differs from memory summaries. - Claude release notes. Claude Help Center. https://support.claude.com/en/articles/12138966-release-notes. Accessed 2026-07-06. Freshness reference for ongoing Claude app and Claude Code product changes.