Foxbra is an integrated AI workforce stack. Each product stands on its own, but together they position AI agents as accountable, contract-bound workers — not mere software tools.
Describe the outcome you want rather than the procedure to get there. A crew of AI specialists handles planning, building, checking, and delivery — so you steer intent while the workforce does the work.
Shipping today at v5.0.5, Studio runs on desktop and web and is the front door to the rest of the stack.
A self-hosted agent runtime built on a single Postgres engine and licensed under Apache-2.0. The Runtime governs autonomous agents the way you'd manage human employees — through intent, authority, evidence, and settlement.
Every job must carry all four before it runs, and prove all four when it completes. Pre-1.0 and open for you to run yourself.
An open specification that contracts what agents may do, what they owe, and what they can prove. Where protocols like MCP and A2A describe an agent's capabilities, GRIP goes further — defining explicit limits, obligations, and proof.
Limits set what an agent may do. Obligations set what it owes, such as settlement receipts within token limits. Proof is content-addressed evidence anyone can verify.
The world's first Agentic Language Model — trained natively against governance contracts rather than conversational patterns. Fennec understands authority, evidence, and settlement as first-class concepts, not afterthoughts.
Ready today and awaiting public release.