What Relay.app does
Relay.app is a collaborative workflow automation platform combining app triggers, deterministic steps, AI models, agentic tool use, and human-in-the-loop reviews.
Relay.app's distinguishing feature is that human work is modeled inside the automation. A reviewer can approve, edit, re-prompt, stop a run, enter missing data, complete an offline task, or select a path, with notifications through email or Slack.
Usage is measured through completed steps and AI credits. Built-in Relay credits can call supported OpenAI, Google, and Anthropic models, while advanced users can connect provider accounts. Plan allowances and prices change with selected usage, so calculate triggers, app actions, human steps, retries, and model calls from the current calculator.
An approval step only helps when it is assigned to the right person with enough context and a safe default. Teams should avoid auto-continuing overdue sensitive reviews, constrain agentic tool sets, protect connected credentials, validate structured outputs, and monitor duplicates, partial failures, loops, and downstream side effects.
How Relay.app works
A trigger starts a workflow and passes structured data through connected app actions, conditions, transforms, waits, and AI steps. AI can summarize, extract, prompt a selected model, or choose among granted tools. Approval, data-input, task, path-selection, and AI-review steps pause the run for a person before downstream actions continue.
How to set up Relay.app
Map one repeated process
Draw the trigger, authoritative inputs, deterministic actions, AI judgment, human decision, failure path, and measurable completion state.
Connect least-privilege apps
Use a test workspace and narrow accounts, fields, folders, and actions; avoid production write access until the run is predictable.
Define structured AI output
Specify required fields, types, examples, uncertainty handling, and validation rather than passing unconstrained prose downstream.
Insert meaningful review
Assign an accountable person, show source context, allow edits or termination, set reminders, and choose a safe overdue behavior.
Test failures and cost
Simulate missing data, invalid output, duplicate triggers, API outages, rejected approvals, retries, and high volume before enabling production.
Relay.app FAQs
What kinds of AI steps does Relay.app support?
Its docs describe built-in AI actions, prompt-any-model steps with structured output, and agentic tool use.
Can a person review AI output?
Yes. Reviews can approve, edit, re-prompt, or end the run, and can be always-on or conditional.
How is Relay.app usage counted?
Plans meter completed workflow steps and AI credits. Triggers and some control steps are excluded, while app actions and activated human steps generally count.
Must users bring model API keys?
No. Relay credits can access supported models, while direct provider connections are available for advanced configurations.
Can approvals prevent every bad action?
No. Reviewers can miss errors or lack context. Use deterministic validation, least privilege, safe defaults, logs, and rollback in addition to approvals.
Listing reviewed 2026-07-23. Product details and pricing can change; verify important terms on the provider's website.
Related Automation AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.