# WhyKit

> WhyKit is an open-source MCP server for developer tools. Read-only access to a Git-based ledger of decisions and which code changes touch them. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI
- **Website:** https://whykit.cometweb.io
- **Repository:** https://github.com/cometweb-io/whykit
- **MCP Registry name:** io.github.CometWeb-io/whykit
- **Categories:** developer tools, knowledge and memory
- **Listing updated:** 2026-11-14

## About

WhyKit keeps decisions and their sources as plain Markdown in the repo. Each decision cites sources by ID, declares the code paths it governs, has an owner and a review date, and is append-only: you supersede a decision rather than rewrite it. A command such as whykit affected takes a list of changed files and reports which decisions apply, which need a second look because a cited source was retired or went stale, and which files no decision covers. It states what it checked and never claims a change is safe.

The MCP server exposes the same vault over stdio or loopback HTTP, read-only, with tools named query, context, impact, status, pack, trace and backlinks. A sensitivity ceiling, defaulting to internal, filters what an agent may see, and the test suite checks the vault bytes do not change. whykit connect tells agents to ask before editing.

For pull requests, whykit check runs locally or as a GitHub Action and fails when a decision still in force rests on something that no longer holds. An optional static Explorer site renders the same files. It is Apache-2.0 licensed with no runtime dependencies and needs Python 3.11 or newer.

## Features

- **Change-to-decision mapping:** Lists which decisions a set of changed files touches.
- **Staleness checks:** Flags decisions whose sources were retired or whose review date has passed.
- **Read-only MCP:** Agents look up decisions without being able to change a file.
- **Sensitivity ceiling:** Limits what an agent can read by record sensitivity label.
- **CI gate:** A check command and GitHub Action fail pull requests that rest on dead evidence.
- **Plain Markdown vault:** Everything is files in Git, openable in Obsidian.

## Good for

- Warn an AI coding agent before it undoes a deliberate design choice
- Fail a pull request when a decision rests on retired evidence
- Trace why a payment or webhook rule exists
- List decisions due for review within a month

## Install

### Claude Code (Terminal)

```sh
claude mcp add whykit -- uvx whykit
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "whykit": {
      "command": "uvx",
      "args": [
        "whykit"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "whykit": {
      "command": "uvx",
      "args": [
        "whykit"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "whykit": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "whykit"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "whykit": {
      "command": "uvx",
      "args": [
        "whykit"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "whykit": {
      "command": "uvx",
      "args": [
        "whykit"
      ]
    }
  }
}
```

## Tools

- `query`: Searches decisions and evidence in the vault.
- `context`: Returns context for a topic or change.
- `impact`: Shows which decisions a code change touches.
- `status`: Summarises the vault: counts, lint results and review queue.
- `pack`: Bundles related records for an agent.
- `trace`: Traces a decision to its sources and predecessors.
- `backlinks`: Lists records that link to a given record.

## FAQ

### Can the MCP server edit my vault?

No. It is read-only and the test suite verifies the files are unchanged.

### What licence is it under?

Apache-2.0.

### What does it need to run?

Python 3.11 or newer and uv, with no runtime dependencies.

### Does it work offline?

It is described as offline by design, and Explorer is a static site.

### Is WhyKit free?

Yes. WhyKit is open source, so you can use and self-host it for free.

### How do I add WhyKit to Claude Code?

Run this command in your terminal: claude mcp add whykit -- uvx whykit

### What are the best alternatives to WhyKit?

The closest alternatives to WhyKit are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Nautilus Compass Memory](https://ai-powered.com/mcp/nautilus-compass-memory): Nautilus Compass Memory is an open-source MCP server for knowledge and memory.

---
Source: https://ai-powered.com/mcp/whykit
