Relay.app

Relay.app is a no-code workflow automation platform with one feature it has chosen to build its whole identity around: native human-in-the-loop approvals. Plenty of tools can connect apps and move data, and several can ping you when something needs attention. What Relay.app does that the giants do not is let a workflow genuinely pause, wait for a person to approve, revise, or reject, and then resume from exactly where it stopped. For operations, finance, and compliance teams who want AI automation but cannot let it fire unsupervised, that pause-and-resume gate is not a nice-to-have — it is the entire reason the category exists for them.

Mechanically, Relay.app looks like a modern automation canvas. You build workflows that trigger on events across your connected apps, add AI steps powered by GPT, Claude, or Gemini for jobs like summarising, classifying, translating, generating images, or transcribing audio, layer in conditional logic, and — crucially — insert approval steps wherever a human judgment is needed. When a workflow hits an approval, Relay.app sends the relevant output to you in email or Slack; you approve it, edit it, or send it back, and only then does the automation continue. Zapier and Make can dispatch an approval notification, but they lack this native pause-and-resume tied to the human decision, which means teams have historically hacked it together with fragile workarounds. Relay.app makes it a first-class primitive.

Pricing is positioned to undercut on entry. There is a free plan for a single user (a few hundred automated steps plus a monthly AI-credit allowance), and paid tiers that have shifted over 2026 toward a per-run model — broadly, a Pro plan in the $10-20/month range and a Growth/Team plan around $30-69/month, with runs counted per workflow execution rather than per step. That per-run accounting is a quiet advantage: a workflow with fifteen steps still counts as one run, so complex automations do not punish you the way per-task billing would. The flip side is that high-volume operations will still climb tiers, and the step and credit ceilings on lower plans are real limits to plan around.

The platform’s February 2026 Agents launch reframed the product. The homepage now leads with “build a team of AI agents” rather than “build workflows,” a deliberate move out of the Zapier-lite comparison and toward the agent conversation. The substance behind the rebrand is the same human-gated philosophy: agents that can take real action, but with approval checkpoints wired in so a person stays in control of anything consequential. That is a sharper, more defensible position than competing on raw automation breadth, where Relay.app cannot win.

And breadth is exactly where it cannot win, which is the honest centre of any buying decision. Relay.app supports roughly 100 integrations. Zapier supports more than 7,000; Make, more than 2,000. If your automation depends on a long tail of niche SaaS tools, there is a real chance Relay.app simply does not connect to one of them, and that gap can be disqualifying regardless of how good the approval gates are. The platform is betting that depth of control on the integrations that matter most beats shallow coverage of everything — a reasonable bet for its target buyer, but one that leaves it unsuitable for sprawling, long-tail automation estates.

Against the tools PixlRun covers, the closest neighbour is Lindy, the no-code agent platform we also reviewed this week, and the contrast is instructive. Lindy is agent-first: you describe an assistant and it runs ongoing processes like inbox management largely autonomously, billed on a credit meter that can spike. Relay.app is workflow-first with agents layered on: you design an explicit, visible automation with human checkpoints, billed per run for more predictable costs. If you want autonomy and are comfortable supervising a credit bill, Lindy leans your way; if you want explicit, auditable workflows where a human signs off before anything fires, Relay.app is the better-fitting tool. The two are converging on the same “AI agents for operations” market from opposite directions.

The verdict: Relay.app is the cleanest answer on the market for AI automation that must keep a human in the loop, and for compliance-sensitive teams that single capability can outweigh everything else. The per-run pricing is sensible, the free tier makes it low-risk to trial, and the approval-gate primitive is genuinely better implemented here than in the incumbents. The reasons to look elsewhere are coverage and maturity: with around 100 integrations and a smaller community than Zapier or Make, it is the wrong pick for broad, long-tail automation. Match it to its purpose — controlled, AI-assisted workflows with human sign-off — and it earns a solid place; ask it to be a do-everything connector and its limits show quickly.

A closing word on fit. Relay.app rewards the buyer who knows exactly which decisions in their process need a human and which do not, then wires that judgment directly into the automation rather than bolting approvals on after the fact. For a finance team approving refunds, a marketing team signing off on AI-drafted copy before it publishes, or an operations lead gating an outbound action behind a quick Slack approval, the pause-and-resume model turns AI automation from a leap of faith into a controlled, auditable process. That is a narrower promise than “automate everything,” but it is a promise the platform actually keeps, and for the teams who need it, a tool that does one important thing properly beats a sprawling connector that does the risky part badly.