The runtime is open
Why we did it
Everything we build runs on one engine — our own company included. That engine is public, under the Apache-2.0 license: read it, run it on your own infrastructure, take it, change it, build on it. We did this for three reasons. Trust: a small company can be deep tech an enterprise relies on only if you don't have to take our word for what the engine does. Honesty: we say the AI has no tools and money moves only through verified flows — the code is how you check. And a foundation for others: the runtime is the on-ramp for builders who will one day serve their own clients on our platform, and the backbone everything we operate stands on.
What it is
An open-source runtime for MCP tools, agents, and workflows — the server that runs them: where tools live, who may call them, what happens when one needs OAuth, a schedule, or a human sign-off. Permissions, audit, and delegated access are part of the open core, because governance is the product, not the paywall.
Tai42 builds and runs its business on this runtime; the code is open; the company sells the hosted platform and enterprise layer on top — never a different core.
What we promise about it
The license stays what it is — no relicensing, no moving the line. Nothing we call open is anything less than open. The runtime we publish is the runtime we run.
How we run it
We answer issues weekly, in batches, because our engineers' first job is shipping client systems on this same runtime — that's the point: this isn't a showcase repo, it's the thing we bet our own delivery on. Questions and proposals: GitHub Discussions. Install from PyPI; start at the docs.