Skip to content

AgentAnywhere Open Source

We open-source the parts you would otherwise have to take on faith.

AI is being built faster than it can be governed. We think the answer is to make AI verifiable — and the components that make governance checkable should not be locked inside a vendor. So we are opening our control points to the community. This is where their documentation lives.

  • Shuddhi


    Training-data provenance. Turns a corpus into a reproducible, verifiable receipt — and emits the EU AI Act Article 53(1)(d) summary.

    Apache-2.0 · published · GitHub

  • Agent Universal Gateway


    The control plane every agent call passes through — nine protocols, one policy surface.

    In preparation

  • Model Hub


    Model identity, versioning and lineage — the record of what is actually in production.

    In preparation

  • Annotator


    The tooling that labels the data the whole chain rests on.

    In preparation


Each of these sits at a control point — what a model may learn from, what an agent may call, which model is serving, and how the data was labelled. A governance claim you cannot read is a claim you cannot check, which is exactly why these are the ones we open.

The wider AgentAnywhere platform — sovereign deployment, the signed audit chain, SSO and multi-tenancy — is commercial. Read the whole boundary on the open-source programme page.