context-harness

Context Harness ingests external knowledge from connectors into a local-first SQLite store (with FTS5 keyword search and optional embeddings), then exposes retrieval via a CLI (`ctx`) and an MCP-compatible HTTP server (plus REST endpoints) for AI tools like Cursor and Claude.

Evaluated Mar 30, 2026 (22d ago)
Repo ↗ Ai Ml local-first rag mcp sqlite embeddings rust microsoft-fts5 cli connectors rag-retrieval cursor claude
⚙ Agent Friendliness
66
/ 100
Can an agent use this?
🔒 Security
38
/ 100
Is it safe for agents?
⚡ Reliability
35
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
75
Documentation
80
Error Messages
--
Auth Simplicity
80
Rate Limits
10

🔒 Security

TLS Enforcement
40
Auth Strength
20
Scope Granularity
20
Dep. Hygiene
55
Secret Handling
65

Strengths: supports local-first storage (SQLite) and local embedding options, reducing data exposure. Uses env vars for OpenAI key in configuration examples, which is generally better than hardcoding. Gaps/uncertainty: README does not describe server-side auth for the MCP/REST endpoints or whether TLS is enforced for the HTTP server; also no mention of access control, request logging/redaction, or rate limiting. Dependency hygiene/CVEs cannot be assessed from README alone.

⚡ Reliability

Uptime/SLA
0
Version Stability
45
Breaking Changes
40
Error Recovery
55
AF Security Reliability

Best When

You want an offline-capable, local knowledge index for AI tooling, with incremental ingestion and MCP/REST endpoints for agent/IDE integration.

Avoid When

You need hosted, multi-tenant, internet-facing APIs with robust server-side access controls; in that case you’ll need to add infrastructure security beyond what’s described here.

Use Cases

  • Local RAG over private files and repositories
  • Indexing and incremental sync of heterogeneous sources (filesystem, Git, S3, custom Lua connectors)
  • Hybrid keyword + semantic retrieval for chat/IDE assistants
  • Using local/offline embeddings for privacy-preserving retrieval
  • Providing tool/agent context to MCP-compatible clients via HTTP

Not For

  • Building a fully managed cloud RAG service (it’s local-first/self-hosted)
  • User-facing production search APIs without additional operational hardening (rate limiting, auth at the server, etc.)
  • Use cases requiring strong enterprise governance features (not evidenced in the provided README)

Interface

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

Authentication

Methods: Environment-variable API keys for embedding provider (OPENAI_API_KEY) Connector-specific credentials configured in TOML for sources like S3/Git (and script connectors)
OAuth: No Scopes: No

No user-auth mechanism for the MCP/REST server is described in the README; integration appears intended for localhost usage and credentialed ingestion via config/env vars.

Pricing

Free tier: No
Requires CC: No

The project itself is open-source; operating costs depend on chosen embedding provider (local/Ollama vs OpenAI).

Agent Metadata

Pagination
none
Idempotent
True
Retry Guidance
Not documented

Known Gotchas

  • Local embeddings may require model downloads on first use, which can increase startup latency for initial indexing.
  • If embedding providers are used (Ollama/OpenAI), availability/credentials affect end-to-end retrieval freshness (sync/embed commands).
  • MCP/REST server appears localhost-oriented in the example; exposing it publicly would require additional security controls not described here.

Alternatives

Full Evaluation Report

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

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