What is ai-memory?
ai-memory is an open-source long-term memory project for AI coding agents and developers who switch between coding tools, machines, or collaborators. ai-memory preserves project knowledge outside an individual agent’s notes, allowing another agent to continue work with information about architecture, unsuccessful approaches, and unresolved questions. The repository distributes ai-memory under the MIT license.
ai-memory uses a git-backed wiki of ordinary markdown files as its source of truth. Developers can search those files with existing tools, open them in Obsidian, or edit them manually. The database serves as a derived retrieval index rather than the only copy of the knowledge, and can be rebuilt from the files.
What ai-memory can do
ai-memory combines automatic work capture, searchable project knowledge, and explicit handoffs between coding agents. Lifecycle hooks record prompts, tool calls, and session boundaries, with observations sanitized before storage. Default capture, search, and handoffs operate without LLM calls or an API key, while optional features add model-assisted consolidation and vector retrieval.
- ai-memory consolidates session observations into readable markdown wiki pages.
- ai-memory searches full text, entities, and links in a combined ranking.
- ai-memory transfers handoffs with an owner and claim-once semantics.
- ai-memory shares project knowledge across users connected to a server.
- ai-memory supplies cross-project profiles of working preferences.
- ai-memory supports reversible memory aging, compaction, and duplicate handling.
How ai-memory works
ai-memory follows a capture, consolidation, recall, and handoff workflow. Agents emit sanitized observations through lifecycle hooks as development proceeds. At session end, those observations become project wiki pages. A subsequent session receives a bounded brief and can search the stored knowledge. Integrations include Claude Code, Codex, Cursor, Gemini CLI, and other coding harnesses.
ai-memory runs through a server hosted on the user’s laptop, a homelab machine, or another location. Starting agents through ai-memory run can also provide a private, temporary local projection of the previous agent’s visible native history. The remote server receives harness and session identity, not the local transcript path or body. Direct agent launches retain their existing behavior.
Who ai-memory is for
ai-memory serves developers who want coding context to survive changes of agent, repository, or machine, as well as teams sharing project knowledge. A shared server makes knowledge retrievable across team members’ agents while keeping personal handoffs personal. Cross-project profiles carry working habits into new repositories and sit below each repository’s own rules.
- Developers can resume unfinished work in a different coding agent.
- Developers can move between desktop and laptop sessions with shared memory.
- Teams can retrieve project knowledge learned during colleagues’ sessions.
- Developers can retain package-manager preferences, test layouts, and architecture habits across projects.
ai-memory pricing and access
ai-memory is available as an open-source project under the MIT license and ships as a self-contained binary. The default workflow requires no API key or LLM calls. Multi-user authentication, per-person attribution, and an audit log are included rather than reserved for a paid tier. Optional LLM functionality requires configuring an LLM.
ai-memory strengths and limitations
ai-memory combines editable markdown storage, cross-agent continuity, and a default workflow without required cloud API usage. Git history and version chains preserve original content when memory is compacted or merged. Access-weighted retention is enabled by default, while content-rewriting or dropping mechanisms remain disabled until enabled. These controls distinguish basic capture and retrieval from optional memory transformation.
- ai-memory requires a user-run server for memory storage and sharing.
- ai-memory lists native Windows support as experimental.
- ai-memory provides MCP-only or hooks-only access for some integrations.
- ai-memory requires build-time and runtime opt-ins for its optional management console.
ai-memory FAQ
What is ai-memory?
ai-memory is an open-source long-term memory tool for AI coding agents. ai-memory captures development activity, consolidates observations into a git-backed markdown wiki, and provides searchable context and explicit handoffs across agents, machines, and teams.
Is ai-memory free?
ai-memory is available under the MIT license, with multi-user authentication, attribution, and audit logging included rather than offered as a paid tier. Default capture, search, and handoffs require no LLM calls or API key.
Which coding agents does ai-memory support?
ai-memory lists integrations for Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Kiro CLI, and other agents. Integration coverage varies: some tools support lifecycle hooks and MCP, while others provide MCP-only or hooks-only access.
Does ai-memory need an LLM or API key?
ai-memory does not require an LLM or API key for default capture, search, and handoffs. LLM-written consolidation and background rewriting of cold notes are optional, and the background rewriting feature is disabled by default.
Where does ai-memory store project memory?
ai-memory stores its source of truth in a git-backed wiki of ordinary markdown files on a user-run server. The database is a derived index that can be rebuilt, and developers can open, search, or edit the markdown files directly.
Can teams share memory with ai-memory?
ai-memory supports teams connecting to a shared server, where agents can retrieve project knowledge learned in other users’ sessions. Personal handoffs remain personal, and multi-user authentication, per-person attribution, and an audit log are built in.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




