Runs

Every time a procedure executes, it creates a run — an immutable record of everything that happened.

What is a run?

Every time a procedure executes, Rapidfolio creates a run — a permanent record of that execution. The run captures everything: the inputs it received, every step that executed with its exact inputs and outputs, every human approval decision, and the final result.

Runs never change after they complete. If you need to re-execute, you start a new run. The original is preserved.

This makes runs the natural audit trail for your financial automation. For any action that happened — a payment, a credit check, a wire transfer, an account closure — you can pull up the run and see exactly what was sent, what came back, and who approved it.

Run lifecycle

A run moves through these states:

pending → running → completed
                 ↘ awaiting_review → running → completed
                 ↘ failed
                 ↘ cancelled

Status	What it means
`pending`	Queued but not yet started
`running`	The AI agent is actively executing nodes
`awaiting_review`	Paused at a Human Review node — waiting for a team member
`completed`	Finished successfully with a final output
`failed`	Stopped due to a tool error, a rejected review, or an unhandled condition
`cancelled`	Manually stopped before finishing

Once a run reaches completed, failed, or cancelled, it's done — it won't change again.

What runs record

Every run captures a complete picture:

The procedure version that executed — you always know exactly which version of your workflow ran
The environment — sandbox or live
All inputs — the exact values passed when the run started
The step log — every node that ran, with its inputs and outputs
Human review decisions — who approved, when, what they saw, and any value overrides
The final output — or, for failed runs, the failure reason and which step caused it

The step log is the most valuable part for debugging and compliance. For each tool call you can see the exact JSON that was sent to the integration and the exact response that came back.

How runs start

Runs can start in three ways:

API — your backend calls POST /v1/procedures/:id/:environment
Trigger — an event fires (Stripe webhook, Linear issue created, cron schedule) and Rapidfolio starts the run automatically
Manual — a team member clicks Run in the dashboard

The run type is recorded on the run record.

Viewing runs in the dashboard

Go to Runs in the dashboard. You can filter by status, procedure, environment, or date range.

Click any run to see:

Overview — status, type, environment, duration
Inputs — the exact values it received
Step Log — every node, in order, with what went in and what came out
Output — the final result (or failure reason)
Events — timeline of status changes, review decisions, and webhook deliveries

Retrying failed runs

When a run fails due to a transient issue — a network error, rate limit, or integration downtime — you can retry it from the dashboard. Rapidfolio creates a new run with the same inputs. The original failed run is preserved unchanged.

Runs that failed because a reviewer rejected them cannot be retried this way. If appropriate, start a fresh run.

Runs and compliance

Every run is permanent and append-only. Rapidfolio doesn't delete or modify run records. For financial operations teams with audit requirements, run records provide a complete, tamper-evident history of every automated action — what happened, in what order, authorised by whom.