MCP Memory Graph
Give your agent a persistent authority-weighted memory graph with semantic search and conflict detection across long projects.
Overview
MCP Memory Graph is an MCP server for the Build phase that provides an authority-weighted memory graph with semantic search and conflict detection for AI agents.
What is this MCP server?
- Authority-weighted memory graph for agent sessions
- Conflict detection when stored facts disagree
- Semantic search over graph-backed memories
- stdio MCP via PyPI package mcp-memory-graph
- Version 0.1.1 in published server metadata
- PyPI identifier mcp-memory-graph with stdio transport
- Features authority weighting, conflict detection, and semantic search per description
What problem does it solve?
Long agent threads lose decisions and silently overwrite facts when multiple sessions disagree on how your product works.
Who is it for?
Solo builders running multi-session agent projects who want graph memory with search and conflict signals via MCP.
Skip if: Teams that only need ephemeral chat context or a full managed knowledge base with human editorial workflows.
What do I get? / Deliverables
After registration, agents can query and update a structured memory graph that highlights conflicts and ranks trusted memories.
- Graph-backed durable memories across agent sessions
- Conflict signals when stored facts disagree
- Semantic retrieval for architecture and debugging context
Recommended MCP Servers
Journey fit
Build → agent-tooling is the canonical shelf for memory infrastructure you attach while constructing agentic products. Agent-tooling subphase is where memory graphs, tool routers, and context systems live—not a one-off IDE plugin.
How it compares
Graph-based agent memory MCP, not a generic notes app or single-file CLAUDE.md-only workflow.
Common Questions / FAQ
Who is MCP Memory Graph for?
Indie developers building agent-heavy products who need persistent, searchable memory with authority weighting and conflict awareness.
When should I use MCP Memory Graph?
Use it from build through operate whenever sessions must recall decisions, detect contradictory memories, and search semantically across prior work.
How do I add MCP Memory Graph to my agent?
Install PyPI mcp-memory-graph 0.1.1, add a stdio MCP server entry in Claude Code, Cursor, or compatible clients, and restart the agent.