Research pass included MCP because agent builders are converging on tool-mediated memory and context access.
Tool profile · provisional profile
MCP memory servers
Promising interface layer. The core question is who can write, what gets stored, and when memory enters the working context.
Provisional fit
68/100
Best for: Teams standardizing memory access as a tool rather than as hidden model-side state.
Avoid if: you need a fully governed, citation-complete knowledge architecture without adding policy, evidence capture, and review workflow around the tool.
Caution: Protocol shape does not solve memory quality. Server policy must handle source, scope, freshness and deletion.
Model signature: Governance primary · tool scope · Protocol/tooling
Layer coverage
Where this tool fits.
This is not a completed review. It is a provisional profile from public positioning plus known failure-mode mapping. Hands-on benchmarks, source snapshots, and citation-bound claims are still required before stronger conclusions.
Production
Curation
Storage
Context Assembly
Governance
Review rule: a tool does not get credit for a layer unless it exposes inspectable behavior, not just a marketing claim.
Evidence notes
What the provisional profile has applied so far.
Mapped to governance because MCP creates explicit capability boundaries and auditable calls.
Needs comparison across server implementations: write policy, storage substrate, correction UX and context assembly rules.
Review packet
What a complete review must contain.
This page exposes the intended review structure. The current artifact is a profile, not a completed evidence-backed review.
Canonical source
https://github.com/modelcontextprotocol/servers/tree/main/src/memory
Strengths
Teams standardizing memory access as a tool rather than as hidden model-side state.
Limitations
Protocol shape does not solve memory quality. Server policy must handle source, scope, freshness and deletion.
Dimension assessment
Scope, volatility, authority, lifecycle, resource economics, interoperability, and evidence quality must each get a rationale and citations before final scoring.
Open questions
- What can be verified from docs, code, issues, benchmarks, and changelogs?
- Where does the tool fail under stale, contradictory, private, or high-cost knowledge?
- Which claims are vendor claims versus independently observed behavior?
Benchmark critique
No benchmark number is accepted as architectural evidence unless it says which layer it tests and what it misses: lifecycle, scope boundaries, authority, context cost, and governance.
Related systems
Related tools should be connected by evidence-backed edges: competes with, integrates with, implements concept, evaluated by, or has governance gap.
Update history
Provisional profile created. Stale-review detection, source snapshots, and changelog watching are required before this becomes a durable review.