Workspace Agent Coordination Protocol
A formal protocol specification for coordinating autonomous agents in distributed, fault-tolerant systems.
WACP defines the rules by which autonomous agents coordinate work -- what agents say, how they communicate, and the structures they operate within. It is a protocol specification, not an implementation -- deliberately transport-agnostic, implementable over files, HTTP, pipes, message queues, or anything else.
Specification Documents
Core
Primitives
Workspace
Isolated execution contexts with hard boundaries
Envelope
Structured message containers for agent communication
Signal
Lightweight notification mechanism
Checkpoint
Immutable snapshots of state
Task
Work units with dependency graphs and clear ownership
Trail
Append-only audit log of every event
Identity
Identity and addressing for agents
User
Human participant representation
Mechanisms
The Five Questions
Where do agents work?
Workspaces -- isolated execution contexts with hard boundaries.
How do agents communicate?
Envelopes and signals -- structured, explicit messages.
How is progress recorded?
Checkpoints -- immutable snapshots of state.
How is work organized?
Tasks -- dependency graphs with clear ownership.
What happened?
The trail -- an append-only audit log of every event.
Resources
Creative Commons Attribution-ShareAlike 4.0 International