Today we are launching a local-installable Developer Edition of Supercog, available to anyone on GitHub.

This is a very cool package that gets you running with the whole platform, locally, in about 10 minutes. Everything runs in Docker on your machine. Hat tip to Declan Brady who did most of the heavy lifting.

The prerequisites are minimal:

• Docker desktop
• Your an OpenAI API key

That's it! We even run a local MinIO server to handle file storage for you (you can use your own S3 bucket if you prefer).

Local advantages

Running agents locally on your machine offers some interesting benefits:

• We enable some tools like our Native Code Interpreter which we can't run safely yet in our cloud version. You should try out the native interpreter - let the LLM run code locally on your machine, including installing new packages!
• Some tools like the Playright tool basically only operate on localhost.
• All your data stays local. We don't see any of it. If you run a local LLM via Ollama then your agent never even interacts with a cloud LLM like GPT or Claude.

Build your own tools

One of the big selling points for Supercog is extensibility. Adding your own custom tools suddenly gives your Agents new powers. You can connect your agent to your custom app API, or give it access to some proprietary data source that you use at work. We think there's a great model where developers at your company write some custom tools which are then "AI enabled" for everyone else in the company.

Writing new tools is really, really easy. If you have any questions about how to do it please reach out and ask.

From dev to production

This package isn't really production ready, but it will be very close to what a full production setup will look like (assuming you don't want to use our cloud version). The main work is to decompose the docker-compose so that you run your own Postgres and Redis servers, and you configure the platform with proper production S3 buckets.