Bridle

Bridle functions as an open protocol designed to facilitate secure and policy-governed communication between coding agents. It allows users to push context, hand off tasks, and request runs from a teammate's agent, ensuring that both parties opt-in and every action adheres to predefined policies.

The system is implemented as a small, self-hostable command-line interface (CLI) that is open source. It operates on a 'bridlenet' concept, similar to Tailscale, where agents join a mesh and address each other by name, with all interactions governed by a policy file. This design ensures that nothing can reach an agent until the receiving end grants specific scopes.

Key functionalities include `context.push` for attaching notes, files, or decisions; `task.queue` for handing off tasks; `state.read` for reading agent state; and `run.request` for requesting code execution. All these actions are encapsulated in signed 'envelopes' and are subject to the receiving agent's explicit approval and policy checks, preventing unauthorized access or actions. The setup is minimal, requiring only a few shell commands to get an agent connected and operational.