knowledge-mcp

knowledge-mcp is a Python MCP server plus CLI for creating, managing, and querying local knowledge bases backed by LightRAG (hybrid vector + knowledge graph retrieval). Users configure an embedding/LLM provider (e.g., OpenAI), ingest documents into per-KB directories, and query the resulting index via an MCP (FastMCP) interface for use by MCP-capable AI clients.

Evaluated Mar 30, 2026 (21d ago)
Repo ↗ Ai Ml mcp rag lightrag vector-search knowledge-graph cli python fastmcp local-first
⚙ Agent Friendliness
52
/ 100
Can an agent use this?
🔒 Security
39
/ 100
Is it safe for agents?
⚡ Reliability
25
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
70
Documentation
60
Error Messages
0
Auth Simplicity
80
Rate Limits
5

🔒 Security

TLS Enforcement
50
Auth Strength
20
Scope Granularity
10
Dep. Hygiene
55
Secret Handling
70

MCP server access/authentication is not described; intended usage appears local/container-based. Secrets for upstream providers are referenced via .env and ${ENV_VAR} substitution, which is a positive sign. TLS/auth for the MCP interface itself is not addressed in README; ensure the MCP transport and network exposure are handled securely when deploying beyond localhost. Dependency hygiene is unknown from provided content.

⚡ Reliability

Uptime/SLA
0
Version Stability
35
Breaking Changes
40
Error Recovery
25
AF Security Reliability

Best When

You want a local MCP-based RAG tool that an AI desktop/IDE client can query, with per-knowledge-base retrieval configuration and custom formatting.

Avoid When

You need a fully managed, internet-accessible service with strong built-in access control, SLAs, and clear rate-limit guarantees.

Use Cases

  • Domain-specific Q&A over a curated document collection (technical docs, internal manuals)
  • Agent workflows that need structured retrieval (entity/relationship + vector hybrid) from a local KB
  • Local knowledge bases for privacy-sensitive RAG use cases where documents stay on the host
  • Consistent response formatting for agents via per-KB configurable user prompts

Not For

  • Public multi-tenant deployment without additional security controls
  • Use cases requiring a standardized hosted API with built-in rate limiting and auth
  • Applications that need rich webhooks or push-style updates
  • Environments where OpenAI API keys cannot be used/managed

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: API keys for upstream LLM/embedding providers via environment variables (e.g., OPENAI_API_KEY) used by LightRAG
OAuth: No Scopes: No

No user/client authentication for the MCP server itself is described; access appears intended to be local/controlled by how the process is started (e.g., Docker volume/container).

Pricing

Free tier: No
Requires CC: No

The package itself is MIT-licensed; external provider usage likely incurs variable API costs.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • MCP server appears to be launched as a CLI command (uvx/uv/docker). Agents must ensure the correct config/base args and mounted knowledge directory.
  • Indexing/ingestion uses external LLM/embedding providers; long-running operations may require timeouts/retries on the client side even if not documented.
  • Per-KB config (mode/top_k/token limits/user_prompt) affects retrieval/response; inconsistent config can lead to unexpected results.

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for knowledge-mcp.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered