Kia MCP Server

A powerful, local MCP server for context augmentation in coding agents/IDEs. Provides repository indexing, documentation search, package exploration, and web research capabilities. Tech stack agnostic, supporting any programming language and framework.

GitHub Repository

Tell your coding agent: "Use kia_package_search_grep to find how error handling is implemented in the requests Python library" or "Search the numpy package for array manipulation examples".

Try indexing public documentation or a repository:

• "Index [https://docs.python.org/3/]"
• "Index [https://github.com/browser-use/browser-use]"
• "Use kia_deep_research_agent to compare the best GraphRAG frameworks and then index the one with least latency."

Check your indexed resources:

• "List my resources" or "Check the status of my indexing jobs"
• Visit your local setup to see all your indexed content.

Improves agent performance by up to 27% through semantic search and Tree Sitter-powered chunking.

Features

• Repository Indexing: Clone and index GitHub repos for semantic search.
• Documentation Search: Index and query web documentation for quick access.
• Package Exploration: Search local or remote packages (PyPI, NPM, Crates.io, Go modules) with regex, semantic queries, and file reading – no indexing needed for remote packages.
• Web Research: Perform web searches and deep multi-step research.
• Codebase Visualization: Generate import graphs and share context across agents.
• IDE Integration: Seamless setup with Cursor, VS Code, Claude Code, and more.
• Performance: Improves coding agent productivity by up to 27% through efficient semantic search and local indexing.
• Free & Open-Source: No subscription required, enhanced privacy with local processing.
• Bug Reporting: Submit issues directly to the GitHub repo for quick feedback.

Quick Setup

Prerequisites

• Python 3.9+
• Parallel.ai API key (sign up at Parallel.ai)

Installation

1. Install prerequisites:

   • uv: curl -LsSf https://astral.sh/uv/install.sh | sh
   • Rust: curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh (source ~/.cargo/env)
   • ripgrep: sudo apt install ripgrep (Ubuntu/Debian) or brew install ripgrep (macOS)

2. Install the server:

   git clone https://github.com/Ash-Blanc/kia-mcp.git
   cd kia-mcp
   uvx --from pyproject.toml sync

   Or as a one-liner (after installing prerequisites):

   git clone https://github.com/Ash-Blanc/kia-mcp.git && cd kia-mcp && uvx --from pyproject.toml sync

3. Get API key:

   • Sign up at Parallel.ai and set export PARALLEL_API_KEY="your_key"

Running the Server

uvx --from pyproject.toml run kia-mcp

Or directly:

uvx --from pyproject.toml run python server.py

IDE Integration

Cursor

1. Install FastMCP CLI: pip install fastmcp
2. Add to Cursor: fastmcp install cursor
3. Restart Cursor. The server will auto-connect.

VS Code

1. Install FastMCP CLI: pip install fastmcp
2. Add to VS Code: fastmcp install vscode
3. Reload VS Code window.

Claude Code

1. Install FastMCP CLI: pip install fastmcp
2. Add to Claude Code: fastmcp install claude_code
3. Restart Claude Code.

Claude Desktop

1. Install FastMCP CLI: pip install fastmcp
2. Add to Claude Desktop: fastmcp install claude_desktop
3. Restart Claude Desktop.

Manual Configuration

For other clients, add to MCP config:

{
  "mcpServers": {
    "kia": {
      "command": "uvx",
      "args": ["--from", "/path/to/kia-mcp/pyproject.toml", "run", "kia-mcp"],
      "cwd": "/path/to/kia-mcp",
      "env": {
        "PARALLEL_API_KEY": "your_key"
      }
    }
  }
}

Usage Examples

Package Search (No indexing required!)

Use package search to find how error handling is implemented in the `requests` Python library: kia_package_search_grep("py_pi", "requests", "error handling")
Search the numpy package for array manipulation examples: kia_package_search_hybrid("py_pi", "numpy", ["array manipulation"])
Read specific file sections: kia_package_search_read_file("py_pi", "requests", "requests.py", 1, 50)

Index Documentation or a Repository

Index public documentation: index_documentation("https://docs.python.org/3/")
Index a GitHub repository: index_repository("https://github.com/browser-use/browser-use")
Use deep research to compare frameworks: kia_deep_research_agent("Compare best GraphRAG frameworks and index the one with least latency")

Monitor Progress & Explore

List your resources: list_resources()
Check the status of your indexing jobs: check_resource_status("repository", "browser-use")

Demo: Analyze a Framework

# Example integration with a coding agent (e.g., Cursor or Claude Code)
# Tell your agent: "Index the NextAuth.js repository, search for JWT patterns, find session docs, and show auth patterns from other repos."
# The agent will use Kia's tools automatically.

Submit Bug Report

Report an issue: kia_bug_report("Indexing fails for large repos", "bug", "Error: timeout after 10 minutes")

Available Tools

• kia_package_search_grep: Regex search in packages (local or remote registries)
• kia_package_search_hybrid: Semantic search in packages (local or remote)
• kia_package_search_read_file: Read package file sections (local or remote)
• index_repository: Index GitHub repositories
• search_codebase: Semantic search in indexed repos
• visualize_codebase: Generate import graph
• index_documentation: Index web docs
• search_documentation: Query indexed docs
• list_resources: List indexed resources
• check_resource_status: Check indexing status
• rename_resource: Rename resources
• delete_resource: Remove resources
• kia_web_search: Web search via Parallel.ai
• kia_deep_research_agent: Deep research via Parallel.ai
• initialize_project: Setup IDE configs
• read_source_content: Read indexed content
• kia_context_share: Share context across agents
• kia_bug_report: Submit bug reports or feedback by opening a GitHub issue

Resources

• server://status: Server status and library availability

Notes

• Indexes stored in /tmp (temporary)
• For production, deploy with FastMCP Cloud

Troubleshooting

• Dependencies not available: Run uvx --from pyproject.toml sync to install dependencies
• API key errors: Check PARALLEL_API_KEY is set
• Indexing fails: Verify git and network access
• IDE not connecting: Restart IDE after adding server
• Package search errors: For remote searches, ensure network access; for local, install packages via pip/npm

For issues, check logs or submit a bug report using the kia_bug_report tool.