- CollectionMemory & State
- Classificationagent-native
- LicenseApache-2.0
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP · REST · CLI · STDIO · http
- Verified2026-10-20
What it does
mcp-memory-service is a self-hosted memory backend for AI agent pipelines: one process exposes REST, MCP, OAuth, CLI, and a dashboard. Agents store decisions, share a causal knowledge graph (typed edges: causes, fixes, contradicts), retrieve in ~5 ms, and run autonomous consolidation — embeddings stay local (ONNX) unless the operator chooses a cloud backend.
Distinct from catalog peers:
| Peer | Difference |
|---|---|
| Zep / Graphiti | Temporal context-graph product vs OSS Graphiti framework. This entry is a multi-transport memory service (REST+MCP+OAuth) with X-Agent-ID |
| Mem0 | Extraction SDK/SaaS. mcp-memory-service is self-host MCP+REST with OAuth DCR |
| Recall | MCP-native self-host memory — narrower transport set, no OAuth/DCR story |
| MemPalace | Palace/drawer metaphor + 44 MCP tools. Different architecture (README compares LongMemEval claims) |
Agent-facing interfaces
- Interaction pattern: CLI + MCP / REST
- Catalog protocols: MCP, REST, CLI, STDIO, http
- MCP: ✅ Available — the service is an MCP memory server (stdio + Streamable HTTP + SSE)
- Agent Skill: ⚠️ No official npx skills add package published yet.
- Onboarding type:
mcp
How agents onboard
Interaction pattern: CLI + MCP / REST
pip install mcp-memory-service
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| PyPI | pip install mcp-memory-service |
| CLI | memory server / memory server --http |
| REST | :8000 — 76 endpoints |
| MCP | stdio + Streamable HTTP + SSE |
| OAuth | 2.0 + DCR for remote MCP |
| License | Apache-2.0 |
Verification
Last verified 2026-10-20.
Sources:
- https://api.github.com/repos/doobidoo/mcp-memory-service