Skip to content
API Circle
Open · Self-funded · Built in the open

An API workspace you can git diff — and an AI can drive

API Circle is an open API client where your collections live in your own Git repo, and a built-in MCP server lets any AI client read, author, and run requests. Desktop, web, CLI, and VS Code — no cloud account, no lock-in.

Start the MCP server
$ npx @apicircle/cli mcp
API Circle — Editor
API Circle request editor showing a request, its response, and the workspace sidebar

Drives any Model Context Protocol client

Claude Desktop Claude Code ChatGPT GitHub Copilot Cursor Continue Cline Zed Windsurf
94 MCP tools in the catalog
17 Auth schemes, all functional
5 Surfaces, one engine
60+ Themes to make it yours
Everything you expect — rebuilt around Git and AI

One workspace, six superpowers

A full-featured API client in the spirit of Postman and Insomnia, with the two ideas the others miss baked into the core.

Git-backed workspaces

Your collections are plain JSON in your own repo — branch, diff, review, merge.

  • Push to save, open a PR from inside the app
  • Auto-created working branches off main
  • Three-way merge with a visual conflict resolver
Learn more

Drive your workspace with AI

A built-in MCP server exposes 94 tools so any AI client can read, author, and run requests.

  • 94-tool catalog: requests, envs, mocks, plans, history
  • Works with 9+ AI clients out of the box
  • Copy-paste config snippet for every client
Learn more

Mock any API in seconds

Point at an OpenAPI, Swagger, Postman, or Insomnia file and get a running mock on localhost.

  • Import OpenAPI / Swagger / Postman / Insomnia
  • Flip a 200 to a 503 to exercise error paths
  • Conditional rules, validation & multipliers
Learn more

17 auth schemes, all functional

Bearer to AWS SigV4 to the full OAuth2 grant set — verified against the relevant RFCs.

  • OAuth2 (all grants, PKCE, device flow, refresh)
  • AWS SigV4, Digest, NTLM, Hawk, JWT
  • Folder-level auth that descendants inherit
Learn more

A CLI for pipelines and agents

Run collections, spin up mocks, and import specs from any terminal — no UI required.

  • mock · mcp · import · export · run · workspaces
  • JUnit reporter for CI pipelines
  • Address workspaces by name or by path
Learn more

Edit the same workspace from VS Code

Nine sidebar views and YAML request editing — no webview, no separate sync.

  • Nine TreeViews + YAML editing with CodeLenses
  • Byte-identical commits with desktop & web
  • Mock servers, secret vault & MCP host built in
Learn more
Version control, natively

Your workspace is a Git repo

Collections, environments, and mock definitions are plain JSON pushed to your own GitHub repo on a working branch. Teams collaborate the way they collaborate on code — branches, diffs, pull requests, review.

  • Auto-create a working branch from main, push to save, open a PR
  • A visual three-way merge resolves conflicts in the folder tree, environments, and release ledger
  • Publish fingerprinted releases that linked consumers pin to — private or a public marketplace
Explore Git workspaces
Workspace · Git
The Workspace panel showing GitHub connection, working branch, and releases
AI-native, not AI-bolted-on

Let an AI client drive your workspace

The bundled MCP server speaks the open Model Context Protocol over stdio. Claude, ChatGPT, Cursor, Copilot, and any other MCP client can scan a codebase, propose a collection, generate runnable code, or spin up a mock — all from the chat.

  • 94-tool catalog — request & folder CRUD, environments, assertions, plans, history, mocks
  • A copy-paste config snippet for every supported AI client
  • Every AI write funnels through the same mutation API the UI uses — safe by construction
See the AI integration
MCP · Connect your AI client
The MCP panel showing the workspace mirror and an AI client config snippet
Local mock servers

Mock any API in seconds

Point API Circle at an OpenAPI, Swagger, Postman, or Insomnia file and get a running HTTP mock on localhost. The Hono-based engine handles $ref dereferencing, per-endpoint overrides, conditional rules, validation, and response multipliers.

  • Flip a 200 to a 503 to exercise error paths
  • Mock definitions sync with the team; the runtime stays on your machine
  • The same engine runs in the app, the CLI, and via the MCP mock.start tool
Explore mock servers
Mocks · localhost
The Mocks panel showing a running mock server on localhost
Runs everywhere you do

Pick your surface

One engine, one workspace format, one mutation API behind all of them. Switch surfaces without switching tools.

Desktop app

Day-to-day development on Windows, macOS, and Linux — OS-keychain secrets and local mocks.

Web app

Zero install. Open the browser, and your first workspace is created automatically.

VS Code extension

Edit the same workspace as YAML, right inside your editor — no webview, no separate sync.

CLI

Run collections, spin up mocks, and import specs in CI pipelines and headless agents.

npm packages

Embed the engine, mock server, and MCP host in your own tooling.

How it works

One engine behind every surface

Every write — from the UI, the CLI, or an AI tool call — funnels through a single mutation API in the core. An AI agent can never produce state the UI couldn't have produced, and vice versa.

01

A workspace is JSON

Two documents — a synced collection tree and per-device runtime state. Readable, diffable, yours.

02

Git is the sync layer

Push the synced document to a repo on a working branch. Teammates pull, branch, and merge it like any file.

03

MCP is a first-class driver

The same parsers and mock engine back all surfaces, so an assistant drives exactly what you can.

Bring your APIs under version control

Open the web app in your browser, or grab the desktop build. No account, no setup — your first workspace is created the moment you arrive.