The Problem
AI Agents operating on WordClaw currently rely on a 'polling' model to fetch tasks or monitor status changes (e.g., checking if a payload was approved by a human supervisor). Polling is computationally expensive, wastes LLM contexts, and introduces execution lag.
The Proposal
I have drafted and pushed RFC 0025 to the repository proposing an expansion of the webhooks architecture. By implementing Server-Sent Events (SSE) natively over the Model Context Protocol (MCP), WordClaw can push JSON payload events instantly into an agent's active connection.
This lets agents subscribe to specific system events (like content_item.approved) and execute reactive downstream actions the millisecond the state changes.
Please see the full technical proposal and architecture plan here: doc/rfc/proposed/0025-agentic-webhooks.md.
The Problem
AI Agents operating on WordClaw currently rely on a 'polling' model to fetch tasks or monitor status changes (e.g., checking if a payload was approved by a human supervisor). Polling is computationally expensive, wastes LLM contexts, and introduces execution lag.
The Proposal
I have drafted and pushed RFC 0025 to the repository proposing an expansion of the
webhooksarchitecture. By implementing Server-Sent Events (SSE) natively over the Model Context Protocol (MCP), WordClaw can push JSON payload events instantly into an agent's active connection.This lets agents subscribe to specific system events (like
content_item.approved) and execute reactive downstream actions the millisecond the state changes.Please see the full technical proposal and architecture plan here: doc/rfc/proposed/0025-agentic-webhooks.md.