Relay and Gravity both target the "AI drafts, human approves" pattern. The category framing is different. Relay puts the workflow first and the AI inside a step. Gravity puts the outcome first and approval inside the sentence.

I'm Aryan, founder of Gravity. The interesting comparison here is not who has more connectors. It is how each product thinks about the human approval step that almost every real-world agent needs.

Why I'm writing this comparison

Approval is the thing that breaks most "AI automation" the moment it ships. Without an approval gate, you get auto-sent emails that read robotic, expense categorisations that miss context, or worse. With approval done badly, you get a queue of pending items the human never gets to. The shape of how you model approval matters more than the AI feature list.

What Relay does

Relay.app is an automation platform where human approval is a first-class node. You build a workflow, drop in approval steps, and route them to specific people via a clean web inbox. Trigger options include schedule, webhook, and various SaaS events.

Where Relay shines:

Relay has shipped a thoughtful approval UX. It is the right pick when the human-in-loop step is the centre of gravity for the work.

What Gravity does differently

Gravity does not have an approval node. It has the phrase "queue for me to approve" inside the outcome sentence. The runtime parses that phrase and creates the approval gate without a builder.

"Every weekday at 6pm, summarise today's HubSpot deals closed-won, draft a personal welcome email to each new customer, and queue them for me to approve before sending."

That sentence creates a recurring job with an approval gate. No nodes, no connectors to drag, no separate step. Approval is a behaviour the runtime supports natively. See how to add a human approval step to an agent for the tutorial.

Side-by-side capability comparison

CapabilityRelayGravity
Building modelVisual workflow with approval nodesOne sentence including approval phrases
Approval UXDedicated inbox per userStandard channel (email, Slack), inline
CollaborationMulti-user editing, comments on flowsPer-agent owner, shared via link
AI useAI step inside the workflowAI as the runtime, not a step
ImplementationDrag and drop builderSentence in a text box
PricingPer-seat plus task usageBundled monthly

The category split

Relay is a workflow builder that handles approval well. Gravity is an agent runtime that treats approval as a phrase. The categories overlap on "AI drafts, human approves," and diverge on how the team builds and owns the work.

If your team wants to share workflows the way they share Notion docs, with comments and edits, Relay matches that model. If your team wants each agent to have a single owner who tweaks one sentence, Gravity matches that model better.

How approval is modelled

This is the technical heart of the comparison. Relay's approval node has:

Gravity's approval phrase reads the same context, but lives differently:

Same approval problem, two product shapes. Relay's shape is better for teams with shared queues. Gravity's shape is better for individual operators who already live inside email or Slack.

A 60-second decision framework

  1. Is the approval queue shared across multiple people? Yes leans Relay. No leans Gravity.
  2. Do you want the inbox to be a dedicated app, or your existing channel? Dedicated leans Relay. Existing leans Gravity.
  3. Do you think in workflows or outcomes? Workflows lean Relay. Outcomes lean Gravity.
  4. Do you need per-step collaboration? Yes leans Relay. No leans Gravity.

When Relay is the right choice

When Gravity is the right choice

Common mistakes buyers make

  1. Treating approval as a feature instead of a behaviour. "Does it have approval steps?" is the wrong question. "What happens if I do not approve for three days?" is the right one. Both products handle it; how they handle it differs.
  2. Building for a team you do not have yet. Relay's collaboration features earn their keep at four operators. At one operator, they are overhead.
  3. Skipping the inbox question. Where do approvals land? If it is a separate web app the operator never opens, the queue grows forever. Pick the inbox before the platform.

Frequently asked questions

What is Relay.app?

Relay.app is an automation platform that pairs scheduled and triggered workflows with native human approval steps. It positions as "human-in-the-loop automation" for teams that want AI suggestions plus a person's go-ahead.

Does Gravity support human approval?

Yes. Gravity supports approval gates declared inside the sentence, for example "draft and queue for me to send." Approval is a behaviour, not a separate workflow product.

How is Gravity different from Relay?

Relay is a workflow builder with first-class approval nodes. Gravity is a sentence runtime where approval is one phrase in the outcome description.

Which has better team collaboration?

Relay leans into collaborative editing of workflows with approval queues. Gravity's collaboration model is simpler, an agent has an owner who edits the sentence.

Which fits a solo founder better?

Gravity, in most cases. Relay's approval-centric model gets stronger as the team grows; for a one-person operator, the canvas adds steps the sentence does not need.

Three takeaways before you close this tab

  1. Approval shape. Both products handle it; the modelling is different.
  2. Inbox question. Where approvals land decides whether they actually get done.
  3. Team size matters. Relay rewards bigger teams. Gravity rewards single operators.

Sources