Generate memory from chat history
Generate memory from chat history is a feature on claude.ai that automatically summarizes your past conversations and builds a persistent 'memory summary' that Claude references at the start of every new standalone chat. Rather than starting each session cold, Claude extracts key facts about your preferences, work context, and ongoing projects from your conversation history and maintains them in a structured profile. This means you no longer need to re-explain your background, writing style, technical environment, or recurring goals at the start of each new conversation.
The memory synthesis runs as a background process that updates within 24 hours whenever conversations are created, modified, or deleted. It is separate from Projects: standalone chats outside of Projects contribute to your general memory summary, while each Project maintains its own isolated memory space. You can view, edit, or delete any stored memory at any time through Settings, and you can pause or fully reset memory whenever you choose.
The feature is available on all claude.ai plans (Free, Pro, Max, Team, and Enterprise) across web, desktop, and mobile. Enterprise and Team administrators have additional organization-level controls, and Enterprise owners can disable and permanently purge all memory data across their organization for compliance purposes.
When you’d use it
- ◆Eliminating repetitive context-setting for writers — A freelance technical writer who works with Claude daily no longer has to open every session by explaining their audience, voice guidelines, or preferred style. Claude remembers that they write for Python developers, use the Google style guide, prefer imperative voice, and flag ambiguities with comments rather than direct edits.
- ◆Onboarding new team members to a shared project — A product team using a shared Claude Project has accumulated 8 weeks of sprint decisions, architectural constraints, and known blockers. When a new engineer joins, they can ask Claude about past decisions and receive accurate summaries drawn from project memory, skipping hours of manual briefings.
- ◆Maintaining client context across an enterprise consultancy — A consulting firm with multiple active client engagements keeps separate Claude Projects per client. Claude remembers each client's regulatory constraints, budget parameters, stakeholder preferences, and previously rejected approaches, allowing consultants to pick up exactly where they left off without re-reading old notes.
- ◆Migrating accumulated context from another AI platform — A developer who has spent months building up a detailed profile in another AI assistant wants to switch to Claude. They export their memory from the previous platform, paste it into Claude's memory import tool, and Claude incorporates their preferred frameworks, coding conventions, and workflow constraints without starting from scratch.
- ◆Iterative style and preference learning for content creators — A content creator repeatedly corrects Claude for using corporate jargon and overly long conclusions. Over time, Claude's memory synthesizes these corrections into standing behavioral rules, so future drafts automatically match the creator's preferred concise, plain-language style without requiring repeated correction.
What changed recently
- ◆2025-09 — Memory from chat history launched on claude.ai, initially available to Team and Enterprise plan users. Included incognito chat mode, project-scoped memory, and organization-level admin controls for Enterprise.
- ◆2026-03 — Memory expanded to all claude.ai users including Free and Pro plans. Memory import and export tool also introduced, allowing users to transfer memory summaries between AI platforms. The import feature was explicitly marked as experimental, with the caveat that imported memories may not always integrate correctly.
- ◆2026-03-16 — Anthropic published detailed support documentation clarifying that memory updates within 24 hours of conversations being created, modified, or deleted, and documenting the exact behavior of Pause vs. Reset memory options.
- ◆2026-05 — Beta testing reportedly began for a 'Memory Files' paradigm, which would organize extracted facts into distinct topic-based documents rather than a single compressed summary note. A background consolidation process referred to internally as 'Dreaming' was also reported. These changes were observed in beta and had not been officially released or documented by Anthropic as of the research date.
This is the short version
The full chapter has three worked examples, the common pitfalls, and the workflow that makes it pay — plus the other 84 features, kept current.
Get Claude Master — $97 →