Product
Run Graph
Watch every step, branch, and retry as it happens, with the exact inputs and outputs on each node.
A run is a graph, not a log. Steps that depend on each other draw as a chain; steps that do not draw side by side and execute at the same time. Clicking a node opens its inspector: what went in, what came out, how long it took, and what it cost.
Live, not polled
The graph streams over a single connection and updates as steps resolve.
Shareable
Any run has a permanent link. Paste it into a ticket and whoever opens it sees the same thing you did.
Parallel branches on their own tracks
Concurrency is visible rather than inferred from timestamps.
const run = await agent("triage").start(ticket)
run.on("step", (s) => console.log(s.id, s.status))Measured outcome
An on-call engineer opens the failing run, sees the retry that exhausted its budget on step nine, and reads the tool response that caused it — without adding a single log line.
Product
- Retry PoliciesPer-tool exponential backoff, declared in one file instead of scattered through your handlers.
- Audit TrailEvery prompt, tool call, and token — retained and replayable, exportable as compliance evidence.
- Human in the LoopPause a run, review the state, approve or amend, and resume from the same step.