reasonkit

ReasonKit is a meta-reasoning suite that provides auditable, structured reasoning through ThinkTools and an MCP (Model Context Protocol) server for AI agent integration. The MCP server portion is exposed via the ReasonKit CLI (rk) and can be started with rk serve, enabling external agents to query the reasoning engine, request multi-step analyses, and receive verifiable outputs. This setup is designed to plug into larger AI workflows, allowing you to route prompts through ReasonKit’s think tools (GigaThink, LaserLogic, BedRock, ProofGuard, BrutalHonesty) and to chain results into an auditable reasoning trace for production use. The server coordinates with the memory layer and web components to provide a cohesive, end-to-end reasoning pipeline that can be scaled or containerized as part of broader AI systems.

Once the MCP server is running, you can connect your agents or orchestration layer to the server endpoint to perform operations such as initiating structured reasoning tasks, fetching multi-perspective analyses, validating claims with triangulation, and obtaining formal reasoning traces suitable for auditing and compliance. Use the provided ThinkTools shortcuts and profiles (quick, balanced, deep, paranoid) to tailor the reasoning workflow to your risk tolerance and turnaround requirements, then expose the results to downstream decision-making services or dashboards.

Prerequisites:

• Rust toolchain installed (Rust/Cargo) or access to the prebuilt binaries via the official installer
• Optional: Docker if you plan to run the MCP server in a container

Installation steps (recommended via Rust tools):

1. Install the ReasonKit suite (core tools plus MCP server integration): cargo install reasonkit

2. Verify installation and available commands: rk --version rk serve --help

3. Start the MCP server for AI agent integration: rk serve

If you prefer containerized deployment (Docker):

1. Pull the ReasonKit MCP-enabled image (example placeholder): docker pull reasonkit/reasonkit-mcp

2. Run the container interactively: docker run -i reasonkit/reasonkit-mcp

Prerequisites for container method:

• Ensure Docker is installed and running on your host
• Expose required ports as needed by your orchestration layer

Notes:

• The exact docker image name and tag may vary; consult the official ReasonKit releases for the correct MCP-enabled image
• If you install locally via cargo, ensure Cargo bin directory is in your PATH

Tips and common issues:

• Environment variables: set your LLM provider keys (e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY) as you would for ReasonKit usage; the MCP server will forward prompts to your configured providers.
• Profiles influence the protocol chains used by the MCP server (quick, balanced, deep, paranoid). Choose based on the desired balance between speed and rigor.
• If you upgrade ReasonKit, re-run the installation to ensure all MCP-related components are up to date.
• When containerizing, ensure the host clock is accurate for reproducible reasoning timestamps and auditing.
• For auditing, enable logs/traces at the MCP server level and route outputs to your observability stack.