brain-mcp
brain-mcp is a local (on-device) MCP server plus CLI that imports an individual’s prior AI conversations, builds embeddings/summaries (optional), and exposes cognitive “tools” for searching, reconstructing context, and analyzing thinking history.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Security posture is largely local-first (no cloud accounts/depicted API), which reduces network exposure. However, explicit transport/authz details for the MCP server are not documented (likely runs locally without TLS/auth). Optional summarization dependencies may call external LLMs (per optional dependencies), which could introduce data egress depending on configuration. Dependency hygiene appears moderate from manifest but no CVE status is provided.
⚡ Reliability
Best When
You want local, private “second brain” capabilities inside an AI desktop/client that can connect to an MCP server to query conversation history and summaries.
Avoid When
You require formally specified REST/GraphQL contracts, documented retry/idempotency semantics, or you can’t control local storage/permissions for your conversation archives.
Use Cases
- • Reconstruct where you left off in a project/domain (open questions/decisions/next steps)
- • Search past conversations by meaning (semantic search)
- • Synthesize opinions/views you formed across many past chats
- • Trace how your thinking changed over time (trajectory)
- • Surface unfinished threads across domains
- • Provide quick stats/coverage/dashboard about what’s indexed
- • Assist with re-entry briefs after being away from a domain
Not For
- • Storing or serving sensitive conversation history to third parties (it’s designed for local use, not sharing)
- • Use as a cloud SaaS memory platform with centralized accounts
- • Workflows requiring strong authenticated multi-tenant access control over a network API
- • Operations where you need guaranteed idempotent read/write semantics (the tool behavior for writes is not described)
Interface
Authentication
README emphasizes 100% local/no accounts. No explicit authentication mechanism, tokens, or authorization model is documented for tool access.
Pricing
No pricing described; it appears to be self-hosted/local with optional use of external LLMs for summarization only via optional dependencies.
Agent Metadata
Known Gotchas
- ⚠ README instructs assistants not to show tool names/technical details, but that’s guidance rather than a contract.
- ⚠ Progressive capability depends on whether summaries were generated; agents may see different output quality depending on indexing state.
- ⚠ Local sync behavior (mtime checks every ~60 seconds during tool calls) can change results mid-session if new files appear.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for brain-mcp.
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.