agent-company-core

An alpha Python framework for durable, policy-aware multi-agent workflows. Bring your own model, keep missions in durable storage, and make the boundary between a model decision and an application effect explicit.

python -m pip install agent-company-core
agent-company demo

The demo is deterministic and offline. It exercises mission persistence, capability delegation, typed decisions, audit events, approval, and STOP.

Start here

• Quickstart for the first library mission.
• Provider matrix for optional model adapters.
• Examples for complete small workflows.
• Architecture before changing the core.
• Security before enabling consequential effects.