Local-first semantic memory system for AI agents with MCP, FTS5, KG, temporal reasoning, and multi-agent sync
This repository also configures its own agents. See what Agentic-Memory tells them →
Skill Claude Code needs its repo
Maintainer procedure for adding a new Claude Code lifecycle hook to the agentic-memory system. Use when an event (session start, pre-tool, post-tool, session end) should automatically do something the agent shouldn't have to remember to do. Don't use for scheduled background jobs (that's add-a-cron-job).
Skill Claude CodeCodex
Maintainer procedure for adding a new background cron job to the agentic-memory system. Use when you need a recurring task (consolidation, dedup, integrity check, etc.) that runs on a schedule. Don't use for one-shot CLI tools (just write a script and document it in docs/how-to/).
Skill Claude CodeCodex
Maintainer procedure for adding a new MCP tool to the agentic-memory system. Use when the user-facing tool surface needs a new verb (e.g., a new read op, a new maintenance op, a new admin op). Don't use for adding CLI scripts (that's add-a-cron-job) or new Claude Code lifecycle hooks (that's add-a-claude-code-hook).
Skill Claude CodeCodex
Maintainer view of the agentic-memory system. Use when onboarding to the codebase, planning a refactor that crosses module boundaries, or answering "what calls what / what depends on what". Don't use for end-user MCP tool questions — that's the user-facing agentic-memory skill.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: