OpenCopilot
Open (formerly Squarelab / OpenChat) · Coding · Productivity · Chatbot
OpenCopilot is an open-source framework that turns your product's APIs into an AI copilot. Feed it an OpenAPI spec and users can run tasks in plain language.
About OpenCopilot
What Is OpenCopilot
OpenCopilot is a self-hosted AI copilot framework built for SaaS teams. It reads your API definitions, works out which endpoint a request needs, fills in the payload, and writes the result back as a chat reply. The whole thing runs under the MIT license, so you can ship it inside a commercial product without paying per seat.
It solves one specific problem: users who don't want to learn your interface. Someone types "open a ticket about my broken export" and the copilot picks your create-ticket endpoint, guesses the title, and calls it. No menu hunting, no help doc. You keep the API you already have and add a natural language layer on top.
The limits matter as much as the features. The repository was archived in March 2025 and the README now says "No longer maintained", so nobody is shipping security fixes. Large APIs need hand-written JSON transformers before the copilot can handle them, and setup assumes Docker plus a developer who's comfortable reading logs.
Getting Started
- Clone the repository and put your key in the
.envfile inside thellm-serverfolder, replacing theOPENAI_API_KEYplaceholder. - Run
make installon Linux ormake install-armon Apple Silicon, then open the console athttp://localhost:8888. - Import your Swagger OpenAPI 3.0 definition and check the actions OpenCopilot validated from it.
- Draw a copilot flow for any task that needs several calls in a row, then paste the widget snippet into your app.
Product Information
A quick look at OpenCopilot's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Backend developers at SaaS companies
- Solo founders shipping an MVP
- Teams with data residency rules
Tasks
- Turning an existing REST API into a chat assistant
- Prototyping natural language actions before a rebuild
- Deflecting repetitive support questions
Scenarios
- Internal admin tools
- Adding a copilot to an early product
- Learning how tool-calling agents work
Key features
OpenAPI Spec Import
You upload a Swagger OpenAPI 3.0 file and OpenCopilot turns each operation into an action the model can choose. It validates the schema first, because a sloppy spec produces bad calls. Bulk import is the fastest path from a documented API to a working AI copilot.
Natural Language API Calls
The copilot reads a request, decides whether an API call is needed, picks the endpoint, and builds the payload from context. Ask for "a new case about a broken export" and the title field gets filled for you. The response comes back as readable text rather than raw JSON.
Visual Flows for Multi-Step Tasks
Simple requests work out of the box. Complex ones don't, and that's what copilot flows are for. You draw the sequence in a visual editor, so the model follows your steps instead of guessing the order of calls.
Embeddable Chat Widget
The widget ships with the repo and mounts with a short script tag or a few lines in your frontend. It's the part your users see, and it keeps the API copilot behind your own login and permissions.
Self-Hosted Docker Deployment
Everything runs through Docker Compose, with Make targets for install, migrations, logs, restart, and a full purge. ARM images cover Apple Silicon laptops. Your API keys, conversation logs, and customer data stay on infrastructure you control.
TypeScript SDK and Backend API
The backend listens on port 8888 and exposes its own endpoints, so you can drive the copilot from a script or another service. A separate TypeScript SDK wraps the same calls for frontend work, which makes the self-hosted AI copilot usable outside the bundled widget.
Pros and cons
Pros
- MIT license with no per-seat or per-message pricing, so cost stays predictable as usage grows.
- Swagger OpenAPI import removes most of the manual work of describing endpoints to a model.
- Self-hosting keeps customer conversations and API credentials inside your own network.
- Copilot flows handle multi-step tasks that a plain prompt-to-endpoint setup gets wrong.
- The repository includes the widget, dashboard, and backend in one place, so there's no glue code to write before a demo.
Cons
- The project is archived and no longer maintained, which means no security patches and no fixes when a dependency breaks.
- Large APIs need custom JSON transformers, so the "import a spec and go" promise holds only for small surfaces.
- Setup assumes Docker fluency, an OpenAI key, and comfort with command line install steps.
- The hosted cloud version that used to sit alongside the open-source repo is gone, so support now comes from reading old issues.
Frequently asked questions
OpenCopilot adds a chat interface to an existing product API. It reads an OpenAPI definition, chooses endpoints based on what the user asks, and returns the result as text inside an embeddable chat widget.
Related content
Explore related tools, skills, and articles for OpenCopilot.
OpenCopilot Alternatives
Udon
Udon · Coding · ProductivityUdon turns an Apple Silicon Mac into a home server you run from a browser, with container controls, file shares, and an optional AI sysadmin.
offstage
offstage cloud · Coding · Productivityoffstage gives AI coding agents and CI jobs a real Mac: a fresh macOS account on Apple silicon, driven over an API and billed by the second.
BotBus
Chengdu Tuyang Tusen Technology Co., Ltd. · Coding · ProductivityBotBus lets you control AI coding agents from your phone and watch: approve commands, start tasks and read results while the code stays on your desktop.