How it works

Plan → tools → result.

The real mechanics behind every run - not a simplified diagram.

1. A trigger starts the run

Every path below funnels into the exact same orchestrator - there's no separate “lite” version for automated runs.

Chat

Type a goal directly and watch it work in real time.

Cron schedule

A standard cron expression - "every Monday at 9am", handled by the same job queue as everything else.

Fixed interval

Every N minutes, for things that don't need calendar precision.

Webhook

An external system hits your endpoint; the run starts the moment it does.

Telegram / Slack / WhatsApp / SMS

A message on a connected channel is a signed webhook too - it starts or continues a run exactly like chat.

2. It plans, then executes - concurrently where it can

Iteration budget

A run defaults to 50 iterations, hard-capped at 200 no matter what's requested - a real ceiling, not just a suggestion to the model.

Parallel tool calls

When one turn asks for multiple tools, they run concurrently, not one after another - a fix that shipped after sequential dispatch once summed durations and blew a run's own time budget.

Anti-runaway limits

Creating agents or pipelines is capped at 3 per run and 15 per hour account-wide - added after a real run once produced 30 near-duplicate agents in a row.

Context management

Once a run's context crosses ~80% of its window, older messages get compressed into a summary - anything you've explicitly pinned survives that compression untouched.

3. Model calls have a real fallback chain

A single model outage doesn't take a run down with it.

1Your own key first, on BYOK - a failure here falls straight through, billed normally after
2Primary model on the platform's own account
3A fallback model on the same account if the primary fails
4A secondary backup relay if both of the above fail
5A last-resort tertiary provider if every Claude path is exhausted

4. A claim-verification pass checks the answer

Before a final answer reaches you, it's checked against what actually happened this run. Phrases like “saved to memory” or “sent the email” are mapped to the specific tool call that would have had to succeed for that to be true. If the claim doesn't match a real, successful tool result, the agent gets sent back once to either actually do it or rewrite its answer honestly - one retry per run, not an infinite argument with itself.

5. It remembers - and grades its own work

Before the run starts

Similar past reflections and any facts you've asked it to save get pulled in by relevance and added to its first turn - it starts each run with the right context, not a blank slate.

After the run finishes

A separate pass asks the model to honestly self-assess: what worked, a 1-10 score, and concrete lessons - stored so the next similar run starts smarter, not from scratch.

6. You stay in control the whole time

Approval queue

Sensitive actions pause and wait - the run polls for your decision rather than assuming a yes.

Budget caps

Checked live, every iteration, against both the run's own cap and your account-wide limit - not audited after the money's already spent.

Scheduling reliability

Scheduled runs retry up to 3 times with exponential backoff if something transient fails, instead of silently not running.