FOXBRA FENNEC · READY, AWAITING RELEASE

The first Agentic Language Model.

Fennec is trained natively against governance contracts, not conversational transcripts. It understands authority, evidence, and settlement as first-class concepts — so accountable behavior is intrinsic to the model, not patched on with prompts.

Conventional language models are trained to be agreeable and fluent. That makes them great conversationalists and unreliable employees — they'll happily claim a task is done without any notion of what “done” provably means.

Fennec is different by construction. Its training objective is the governance contract itself: it reasons about the limits it holds, the obligations it owes, and the proof it must produce. When Fennec finishes a job, settlement is native behavior, not an afterthought.

WHAT MAKES IT AGENTIC

Contracts in, proof out.

Contract-native

Trained against GRIP contracts, Fennec treats limits and obligations as inputs it reasons over — not instructions it may ignore.

Evidence-minded

It plans work so that each step can be proven, producing content-addressed evidence as a natural part of acting.

Settlement-aware

Fennec knows a job isn't finished until it's settled — obligations met, within limits, with signed receipts.

DESIGNED TO RUN

Made for the workforce.

Runs under the Runtime

Fennec plugs directly into the Foxbra Runtime, so its behavior is governed and auditable end to end — no special casing.

Speaks GRIP fluently

Because the model and the protocol were designed together, contracts round-trip cleanly from request to settled receipt.

SPEC

At a glance

Ready today and awaiting public release. Join the early-access list to run Fennec inside your own governed workforce.

Agentic LM Contract-native Ready Awaiting release
Request early access Read the GRIP spec