Automations

In brief: Automations connect project events to catalog workflows so cloud agents fix issues, implement stories, and repair failing tests without waiting for someone to paste a /testchimp prompt.

Automations turn TestChimp workflows into event-driven runs. When something changes in your project—an issue is created, a story moves to ready, a CI batch fails, a release lifecycle status changes—TestChimp can hand the matching work to a cloud agent as a catalog workflow execution.

Local agents (Claude Code, Cursor, and the TestChimp skill) still run workflows on demand. Automations cover the complementary case: the platform watches for events and invokes the same workflows for you.

Key concepts

An automation is:

Trigger + Action + Config

Part	What it answers	Examples
Trigger	When should this run?	Entity (issue, story, scenario, test execution, batch, release) + event (created, status updated, failed, …) + optional AND conditions
Action	What should the agent do?	Execute a catalog workflow (fix-issue, implement, run-qa, …) with a task template, optional policy, and an invocation strategy
Config	How should events be batched and gated?	Aggregation (Immediate / Buffered / Windowed) and optional human approval before invoke or before plan execute

Policies stay *.policy.md support files in your repo. An automation optionally attaches a policy to the workflow action—the same files local /testchimp runs use.

How it fits with workflows

Project event  →  matching automation(s)  →  aggregation window
                                              ↓
                                    workflow execution queued
                                              ↓
                         (optional) human approve invoke / plan
                                              ↓
                    cloud agent via OpenHands or GitHub Issue
                                              ↓
                         /testchimp <workflow> … --workflow-execution-id …

• Workflows define how the agent should work (playbook + policy).
• Automations define when that work should start and how the cloud agent is reached.
• Workflow executions are the auditable runs—status, prompt, OpenHands conversation or GitHub issue, approvals—under Executions → Workflow Executions.

Cloud agent invocation strategies

Each automation action picks one project-level strategy:

1. OpenHands — TestChimp starts an OpenHands conversation with the rendered /testchimp … prompt (and optional branch).
2. Via GitHub Issue — TestChimp opens a labelled issue (testchimp plus any labels you configure). Your cloud coding agent (Copilot coding agent, OpenHands resolver, Codex, and similar) watches those labels and runs the prompt in the issue body.

Configure strategies under Project Settings → Automations. Details: Cloud agents. To wire Cursor, Claude Code, Codex, Copilot, or OpenHands (skill + MCP + secrets): Setting up cloud agents.

Where to configure

Surface	Path
Automations list / create wizard	Workflows → Automations (/workflows/automations)
Project rules (rate limit, OpenHands, GitHub labels)	Project Settings → Automations
Runs & approvals	Executions → Workflow Executions

You can also open the create wizard from contextual Automations CTAs on Issues, Plans, Automation Batches, Releases, and individual workflow detail pages (entity / workflow pre-selected).

Default rate limit: 5 cloud agent invocations per hour per project (shared across all automations). Raise or lower it in project automation settings.

Next steps

• How automations work — Triggers, conditions, templates, aggregation, approvals, and tracking
• Cloud agents — OpenHands vs GitHub Issues setup
• Setting up cloud agents — Skill, MCP, secrets, provider guides
• Create an automation — Step-by-step wizard guide
• Typical setups — Common QA recipes
• Announcement: Put QA on Auto-Pilot with Workflow Automations