The API workspace that lives in your repo — and talks to your AI.
Build, test, mock, and ship APIs the way you ship code:
branches, diffs, pull requests — and any MCP-capable AI client on the keyboard.

API clients haven't kept up. Postman pushed your collections into a SaaS silo. Insomnia bolted AI on as an afterthought. Both leave you syncing data through someone else's cloud, with no real story for the new shape of engineering work: branch-and-PR collaboration and AI agents that need to drive your tools.

API Circle starts from a simpler premise — your API workspace is just code. Store it like code. Review it like code. Let agents read and write it like code. Everything else follows.

API Circle Studio is a desktop, web, and CLI API workspace built on three load-bearing ideas:

Collections, environments, and mock definitions are plain JSON pushed to your own GitHub repo on a working branch. Teammates pull, branch, and merge them like any other file. Per-device runtime state — history, sessions, UI — stays local. Secrets are encrypted on-device with AES-GCM, wrapped by the OS keychain on desktop. Nothing is uploaded to a third-party server, ever.

A built-in Model Context Protocol server exposes 71 tools that any MCP client can drive — Claude Desktop, Claude Code, ChatGPT, GitHub Copilot, Cursor, Continue, Cline, Zed, Windsurf. The catalog covers request and folder CRUD, environment authoring, assertions, execution plans, history, mock-server lifecycle, codebase scanning, imports, code generation, and natural-language authoring. Not a plugin. Not a wrapper. A first-class surface, equal to the UI.

Point API Circle at an OpenAPI, Swagger, Postman, or Insomnia file and get a running HTTP mock on localhost in seconds. The Hono-based engine handles $ref dereferencing, per-endpoint overrides (flip a 200 to a 503 to exercise error paths), conditional response rules, request validation, and response multipliers. Mock definitions ship with the repo. Mock runtime stays on your machine.

• 17 authentication schemes, all end-to-end functional — Bearer, Basic, API key, custom header, the full OAuth2 grant set (client credentials, auth code, PKCE, password, implicit, device flow, with auto-refresh), AWS SigV4, Digest, NTLM, Hawk, and JWT. Signing primitives verified against RFC and NIST reference vectors.
• Imports that work — cURL, OpenAPI/Swagger, Postman collections, Insomnia exports, and HAR files.
• Code generation for any saved request — cURL, fetch, Node (axios), Python (requests), Go, and Rust.
• Environments with priority ordering and cross-workspace variable sources.
• Assertions and execution plans to chain requests, runnable headlessly in CI with apicircle run (text, JSON, or JUnit reports).
• Request history with full headers, body previews, and assertion results.

npx @apicircle/cli mock   ./openapi.yaml              # local mock server from a spec
npx @apicircle/cli mcp    --workspace ./workspace     # MCP server for any AI client
npx @apicircle/cli import ./postman_collection.json   # bring your existing collections
npx @apicircle/cli run    "Smoke Tests"               # run a saved plan in CI (text/json/junit)

• Own your data. A workspace you can cat, diff, and git log is a workspace nobody can hold hostage.
• Git is the sync layer. Branches and pull requests already solved collaboration. We didn't need to reinvent that.
• AI is a first-class user. Agents shouldn't have to scrape a UI. They get the same typed API the UI gets.
• One engine, every surface. Desktop, web, CLI, and embeddable packages share one workspace format and one mutation API.
• Open standards only. MCP for AI, Git for sync, OpenAPI / Postman / Insomnia / HAR for import. No proprietary lock-in.

API Circle Studio is pre-launch and self-funded. Expect rough edges and the occasional breaking change before v1.0 — and a team that genuinely reads every issue.

• ⭐ Star studio to follow progress
• 📦 Grab a build from the latest release
• 🐛 File bugs and ideas in issues
• 💬 Trade notes in the Discord

License — API Circle Studio is released under a custom source-available license: free for personal, educational, and non-commercial use (plus a 30-day commercial evaluation). It is not an OSI-approved open-source license; ongoing commercial use requires a separate license. See LICENSE.

_{Own your APIs. Version them like code. Drive them with AI.}