- CollectionMemory & State
- Classificationagent-native
- LicenseMIT
- MCPUnavailable
- Agent SkillUnavailable
- InterfacesAgent Skill · REST · CLI · http · API
- Verified2026-09-26
What it does
Memmy provides one local-first memory substrate shared by Codex, Claude Code, Cursor, OpenClaw, Hermes Agent, and other agent surfaces. Hooks, plugins, the desktop app, and the memmy-memory CLI all read and write the same SQLite-backed service, so context survives both session restarts and switches between agent products.
Its memory pipeline is more than storage. turn.start decides whether and what to recall; parallel vector, full-text, pattern, and structural channels retrieve candidates; fusion, thresholding, deduplication, MMR, and an optional LLM filter select context. turn.complete captures a raw turn and L1 traces, while background jobs summarize, reflect, score, embed, and evolve higher-level L2 policies, L3 world models, and callable Skills. The full lifecycle is documented in How Memory Works.
Agent-facing interfaces
- Interaction pattern: local daemon + JSON CLI / REST + installed Agent Skill
- Catalog protocols: Agent Skill, REST, CLI, http, API
- MCP: ⚠️ No standalone MCP server is documented for the Memory service.
- Agent Skill: ✅ Available — bundled with the repository and rendered for supported external agents by Memmy; this is not published as an npx skills add package.
- Onboarding type:
agent-skill
How agents onboard
Interaction pattern: local daemon + JSON CLI / REST + installed Agent Skill
git clone https://github.com/MemTensor/memmy-agent.git && cd memmy-agent
cp .env.example .env
bash scripts/dev-start.sh
memmy-memory init
memmy-memory health
memmy-memory search "memory policies in this project"
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| Memory REST API | Local HTTP at 127.0.0.1:18960; health, sessions, turns, search/add/get/delete, panel, and analytics endpoints |
memmy-memory CLI |
JSON output; health, session, turn, search, add/get/delete, config reload, and raw calls |
| Agent Skills / hooks | Installs active memory instructions into external agent surfaces and mirrors turns automatically |
| Agent API | OpenAI-compatible local API on port 18990 when memmy serve runs |
| Desktop / TUI | Human management and local Agent Runtime entry points; not required for each memory operation after setup |
Verification
Last verified 2026-09-26.
Sources:
- https://github.com/MemTensor/memmy-agent/releases/tag/v1.1.8
- https://memmy.bot/
- https://api.github.com/repos/MemTensor/memmy-agent