Skip to main content
MCP server CX Codex

Tracekit MCP gateway — Claude Code MCP server: setup

Tracekit MCP gateway checks agent tool calls, signs each decision, and stores a verifiable record for every run. Use it when the agent’s own log is not enough.

MIT ★ 798 updated 2026-10-24

Open on GitHub

What Tracekit MCP gateway does

Tracekit MCP gateway sits between an agent and its tool calls. A Tracekit signer decides whether each call may run, then adds the decision and call to a hash-linked record. Every run produces a .tkb file that can be checked offline, with verifiers available in Python and TypeScript.

Deploy the gateway first; the agent sends a token with each run. Claude Code connects through claude mcp add, while Codex, Cursor, GitHub Copilot, and Gemini CLI use their own MCP settings. In development mode the signer runs as the agent’s user, so the agent could read the key and sign an altered log; a protected setup places the signer beyond the agent’s reach and uses a witness.

What it helps with

  • check tool calls before execution
  • sign agent decisions into a hash chain
  • save one .tkb record per run
  • verify a run without the agent
  • detect edits to recorded events

Example prompts for the agent

  • Check every tool call before it runs and record the decision.
  • Deny the dangerous call, then show the signed run record.
  • Verify the latest .tkb file offline.

How to install Tracekit MCP gateway

1Template address

The server address is a template: replace {…} or […] with your own values

2Platform CLI: remote server

claude mcp add --transport http tracekit-mcp-gateway https://{tracekit_gateway}/mcp

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Tracekit MCP gateway

How do I install Tracekit MCP gateway in Claude Code?

Add the gateway with claude mcp add --transport http tracekit-mcp-gateway https://{tracekit_gateway}/mcp.

What does Tracekit MCP gateway do?

It evaluates agent tool calls, signs the call and decision, and writes a verifiable .tkb record for the run.

What is required for a protected setup?

Deploy the signer where the agent cannot reach it. Each run also sends its token, and the resulting record can be verified offline.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open