ClawJetty is a small developer tool with a simple premise: when an AI agent starts a task, it should immediately hand the user a link to watch what happens next.
The integration is a three-step loop. An agent calls POST /api/runs with a task title, gets back a viewer URL at clawjetty.com/t/{run_id}, and returns that link to the user before doing anything else. It then posts plain-text progress events — update, complete, or fail — via POST /api/runs/{id}/events as the task proceeds. The public page updates in real time without polling or terminal-watching.
The pattern has clear precedent. CI/CD pipelines have produced shareable build status pages for years, and incident tools like StatusPage built entire businesses around live visibility into system state. ClawJetty applies the same logic to agent runs, treating each task as a browser-readable artifact with its own URL. Because the integration happens at the system-prompt level rather than in framework code, it works with any agent capable of making HTTP requests — Claude, GPT-4o, open-source models alike.
The tool is clearly early-stage and deliberately narrow: no authentication, no paid tiers, no enterprise dashboard. It exists to close one specific gap — between an agent starting work and a user knowing what's happening — without demanding significant infrastructure decisions. For developers shipping long-running agent tasks, that may be exactly enough.