memory-graph
MemoryGraph is an MCP server for AI coding agents that provides persistent, graph-based memory. It lets agents store “memories” (typed items with tags/importance) and create relationships between memories, then retrieve them via recall/search tools (including an “extended” profile with more complex queries and DB statistics).
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Security posture is only partially assessable from README/manifest snippets. TLS is likely assumed for any networked cloud endpoint but not explicitly documented. Auth appears optional (API key for cloud backend) without described scopes/least-privilege. No explicit guidance on logging of sensitive values or safe handling of secrets is included in the provided content. Dependency hygiene is plausible but unverified for CVEs in the supplied text.
⚡ Reliability
Best When
You’re using an MCP-capable coding agent and you want structured, relationship-aware recall beyond flat text or basic retrieval.
Avoid When
You cannot control what the agent stores (or you don’t have a prompting/protocol strategy), and when you need fully documented production SLAs/error semantics beyond what the README shows.
Use Cases
- • Persisting decisions, bug fixes, and code patterns across agent sessions
- • Answering multi-hop questions by causal/relational chains (e.g., what led to a fix)
- • Auditing or reporting on knowledge coverage across a codebase
- • Assisting coding agents with retrieval-augmented memory for long-running projects
Not For
- • A standalone autonomous memory system that saves without agent/tool invocation
- • A generic vector-search service only (this is relationship/graph-oriented)
- • Production deployments where strict enterprise security/compliance needs are not documented
Interface
Authentication
README references an API key for a future/coming cloud backend (memorygraph.dev). No detailed auth flows, scope models, or token handling guarantees are documented in the provided text.
Pricing
Only a reference to a “free API key” signup at memorygraph.dev is mentioned; no pricing tiers/limits are provided in the supplied content.
Agent Metadata
Known Gotchas
- ⚠ MCP tool providers don’t automatically store memories; the agent must be prompted/configured (e.g., via CLAUDE.md protocol) to use tools.
- ⚠ Persistence/behavior differs by backend mode (SQLite vs other backends); ensure correct backend configuration for multi-session expectations.
- ⚠ If relationships are important, the agent must explicitly create them (tool calls) rather than assuming they are inferred automatically.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for memory-graph.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.