Changelogai

What changelogai MCP server does

The changelogai MCP server exposes Git-based release workflow functions through Model Context Protocol. It reads commit history from a repository, generates a changelog, and can suggest whether the next release should be major, minor, or patch. The changelog output supports Markdown, JSON, and plain text formats.

The underlying package is designed to work with conventional commits. It can identify the latest tag automatically, group entries into categories such as Features, Bug Fixes, and Performance, and use a specified commit or tag range instead. These behaviors make it suitable for repositories where commit history is the source of release-note content.

How it works

The changelogai_generate tool creates a changelog from Git history. The result can be consumed directly by an agent or written into a release workflow, depending on how the surrounding client handles the tool response. The changelogai_version_bump tool analyzes commits and returns a suggested semver increment: major, minor, or patch.

Repository selection and commit-range controls are available in the command-line package, including a repository path, starting reference, ending reference, and maximum commit count. By default, the package uses the current repository, the latest tag as the starting point, HEAD as the ending point, and a limit of 500 commits.

Optional AI mode uses Claude to turn technical commit messages into descriptions intended for end users. The default model is Claude Haiku, and this mode requires ANTHROPIC_API_KEY. Without AI mode, changelog generation relies on the Git history and commit parsing behavior provided by the package.

Setup and configuration

Install the package globally with npm install -g @ura-dev/changelogai, or let an MCP client download it through npx. A stdio configuration using npx is:

{
  "mcpServers": {
    "changelogai": {
      "command": "npx",
      "args": ["-y", "-p", "@ura-dev/changelogai", "changelogai-mcp"]
    }
  }
}

If the package is installed globally, configure the client to run changelogai-mcp directly. For AI-enhanced descriptions, provide the ANTHROPIC_API_KEY environment variable to the process. Basic changelog generation and version analysis do not require that key.

Tools and capabilities

• changelogai_generate: Generates a changelog from Git history and supports Markdown, JSON, or plain-text output.
• changelogai_version_bump: Examines Git commits and suggests a major, minor, or patch version increment.
• Conventional-commit parsing and optional category grouping.
• Optional Claude rewriting for user-facing release descriptions.

Limitations and notes

The server depends on access to a Git repository and its history. Its output quality therefore depends on the available commits, tags, and commit conventions. AI rewriting is optional rather than part of every generation request, and it introduces a dependency on Anthropic credentials. The provided tools cover changelog generation and semver recommendations; the material does not describe tools for editing commits, creating tags, or publishing releases.