context-harness

Overview

Context Harness is a comprehensive framework designed for local-first context ingestion and retrieval. It supports multiple data sources including filesystems, Git repositories, S3 buckets, and custom Lua scripts. The system stores data locally in SQLite with optional embeddings for semantic search. It exposes functionality through a CLI tool and an MCP-compatible HTTP server, enabling AI tools like Cursor and Claude to access your context. The architecture supports keyword, semantic, and hybrid retrieval modes with various embedding options including local fastembed, Ollama, and OpenAI.

Try asking AI

After installing, here are 3 things you can ask your AI assistant:

• searchsearch("query string")

  Search indexed documents using keyword, semantic, or hybrid mode

  Try:Search for documentation about authenticationTry:Find all files containing error handling patternsTry:Search for information about S3 connectors
• getget("document id")

  Retrieve a specific document by its ID

  Try:Get the document with ID abc-123Try:Retrieve document 456-defTry:Show me the document with id 789-ghi
• sourcessources()

  List available connectors and their status

  Try:List all available data sourcesTry:Which connectors are configured?Try:Show me the status of all connectors
• embedembed({action: "pending" | "rebuild"})

  Backfill missing embeddings or regenerate all embeddings

  Try:Embed all pending documentsTry:Rebuild all embeddings from scratchTry:Check which documents need embeddings
• exportexport()

  Export index as JSON for static site search

  Try:Export the index for static searchTry:Create a JSON export of all documentsTry:Prepare the data for static site search
• agents_listagents_list()

  List all registered agents

  Try:List all available agentsTry:Show me what agents are installedTry:Which agents can I use?
• agent_promptagent_prompt({name: "agent_name", prompt: "query"})

  Resolve an agent prompt with specific agent

  Try:Use code-review agent to analyze this changeTry:Ask the security agent to scan this codeTry:Get feedback from the documentation agent
• listlist()

  List all registered tools

  Try:List all available toolsTry:What tools can I use?Try:Show me all registered functions

Comparable tools

Installation

# macOS (Apple Silicon)
curl -L https://github.com/parallax-labs/context-harness/releases/latest/download/ctx-macos-aarch64.tar.gz | tar xz
sudo mv ctx /usr/local/bin/

# Linux (x86_64)
curl -L https://github.com/parallax-labs/context-harness/releases/latest/download/ctx-linux-x86_64.tar.gz | tar xz
sudo mv ctx /usr/local/bin/

{
  "mcpServers": {
    "context-harness": {
      "url": "http://127.0.0.1:7331/mcp"
    }
  }
}

ctx serve mcp

On Hacker News

Recent discussion from the developer community.

• Show HN: Context Harness – Local first context engine for AI tools▲ 2💬 4
  Story by __parallaxis · 2026-02-26

Compare context-harness with