Developers
The same PHRONEN, over HTTP
Create work, follow it live, be told when it needs a decision, and read the evidence — over HTTP, from the terminal, or inside Zed.
From an outcome to its evidence
01Hand it an outcome
curl https://api.phronen.ai/api/v1/tasks \ -H "Authorization: Bearer $PHRONEN_KEY" \ -H "Idempotency-Key: release-2026-09-20" \ -d '{ "goal": "Get the site live on the new domain and tell me when it is." }'02Follow it, live
curl -N https://api.phronen.ai/api/v1/tasks/$ID/events \ -H "Authorization: Bearer $PHRONEN_KEY" -H "Accept: text/event-stream" event: stage {"stage":"execute","status":"running"} event: approval {"proposal_id":"proposal-from-this-event","action":"Point DNS at the new host","risk":"medium"} event: evidence {"kind":"command","status":"pass"} event: done {"status":"completed"}03Decide on the proposal you inspected
curl -X POST https://api.phronen.ai/api/v1/tasks/$ID/approve \ -H "Authorization: Bearer $PHRONEN_KEY" \ -d '{ "decision": "approve", "proposal_id": "proposal-from-this-event", "note": "go" }'04Read the proof
curl https://api.phronen.ai/api/v1/tasks/$ID/evidence \ -H "Authorization: Bearer $PHRONEN_KEY" { "task": { "status": "completed" }, "evidence": [ { "kind": "command", "ref": "dig +short example.com", "status": "pass" }, { "kind": "source", "ref": "https://example.com", "status": "pass" } ] }
Small on purpose
One task, shared by the web, the phone, the menu bar, the terminal and Zed. If it happens in the product, it is on the API.
- Tasks
Create, list, follow, approve, pause, resume and cancel work.
- Evidence
What was checked, how, and whether it held.
- Devices
The devices in an environment; revoke one in a call.
- Webhooks
Signed deliveries when work waits, records evidence, or finishes.
- Credits
Balance, reservations, what each task used, budgets.
- Keys
Scoped, hashed at rest, shown once, revocable.
This computer, over the same API
MCP servers, slash commands and tools are configured on a connected computer. They are not a hosted catalog. Keys, webhooks, and phr use the same tasks and evidence.
Where you already are
Start work from a shell script or a CI job, and follow it anywhere. In Zed, PHRONEN works inside the editor with the same approvals and evidence.
# a script, a CI job, your shell
phr run "fix the flaky test and prove it holds across 200 runs" --detach
phr watch $ID
phr evidence $ID
# every event as NDJSON, for anything that reads a stream
phr --stream-json run "…"