Git-native, local-first, ready for AI agents

APIGit started in 2022 from one idea: the best place for API work is the place developers already trust with their code. So every spec, document and mock lives in a real Git repository.

That choice turned out to matter twice. Plain files with a history are what people review well, and they are also what AI agents read and write well.

Today the same repository serves both: your team in the editor and in pull requests, and their agents through MCP and llms.txt. And because a clone holds everything, the work can happen on your own machine, where the apigit command line and its Agent Skill let a coding agent such as Claude Code or Codex do the whole job from your terminal.

~/petstore
$ git log --graph --format="%h %an: %s"*   9be04c7 maya: merge: add pet search by name (#42)|\| * 5d21f0a agent: feat: add GET /pets/search|/* c7f8a90 noah: docs: write the getting started guide* d0e1f23 noah: design: draft OpenAPI 3.1 spec$ apigit doc publish --version 1.2.0 --access public

One history: people and an agent commit, a person merges, the documents follow.

Three decisions shape everything

Each feature of APIGit follows from one of them, and usually from more than one.

Git-native: the repository is the product, not an export of it

Every specification, Markdown page and mock script is a file in a real Git repository. APIGit adds the editor, the documents, the mock servers and the rules on top. It never keeps a second copy somewhere you cannot clone.

  • Clone, branch and push over standard Git, from any client
  • Pull requests with the governance findings beside the diff
  • Documents and mock servers published from a commit you choose

Local-first: the work happens on your machine, with the agent you already use

A clone is a complete workspace. The apigit command line checks, previews and mocks an API on your machine, offline and without an account, and an Agent Skill teaches a coding agent the same commands. So Claude Code, Codex, Cursor or whichever agent comes next can do the whole job in your terminal, and APIGit is where the result is reviewed and published.

  • OpenAPI checks and mocks work offline; preview downloads the viewer once
  • apigit mock serve runs the engine of the hosted mock servers locally
  • An Agent Skill and a one-command MCP setup for coding agents

Ready for AI agents: they read and contribute like the rest of the team

The people who use your API now ask an assistant first, and the people who design it work with one. So every published version answers in the formats agents read, and an agent that changes a spec does it the way a colleague would: on a branch.

  • llms.txt, Markdown and an MCP server for every published version
  • An MCP server for your account, signed in with OAuth, scoped by token
  • Agents commit to a branch and open a pull request; people merge

Four rules we build by

They decide what goes into the product, and what stays out.

A spec is source code

It deserves what code has: history, review, blame and a branch for every idea. A record in a portal has none of them.

Review is where quality happens

OpenAPI rule findings guide reviews in the editor and in the pull request. Breaking-change checks from the CLI in CI help catch changes that can break a consumer.

Agents propose, people decide

An agent can draft an endpoint in seconds. Whether it ships is a merge, made by someone who answers for the API.

Open formats, no lock-in

OpenAPI, AsyncAPI, GraphQL and Markdown in a Git repository. OpenAPI checks run in the apigit command line and in CI, and you can leave with everything.

Our mission: make every API as easy to work with for an AI agent as for the engineer who designed it, with Git as the record of who changed what, and why.

For teams of every size, inside the workflow they already use.

Put your API and docs in a repository today

Create a free workspace, add an OpenAPI file and a Markdown guide, and publish your first documentation site and mock server in minutes.

Free for individuals. No credit card.

$ npm install -g @apigit/cli$ apigit auth login$ apigit doc publish