Relay.app Review: AI-Native Automation With a Human in the Loop
Workflow automation is a crowded market, dominated for years by Zapier and Make. Relay.app enters it with a clear thesis: automation should be AI-native from the ground up, and it should treat human judgement as a first-class part of a workflow rather than an afterthought. After testing the platform and verifying its pricing and capabilities against Relay's own materials in July 2026, we think that thesis is well executed — and that Relay has become one of the more compelling options for operations and revenue teams that have outgrown, or been frustrated by, the incumbents.
This review is aimed at the buyer deciding where to standardise their team's automation. We look at what Relay does differently, where its pricing lands, and the situations where a legacy tool is still the safer choice.
Human-in-the-Loop as a Core Primitive
The feature that most clearly distinguishes Relay.app is its handling of human-in-the-loop steps. In most automation tools, inserting a human approval into an otherwise automated process is awkward — bolted on with a chain of conditional emails and status flags. In Relay, it is a native step type. A workflow can pause, request a specific approval, data input or review from a named person, and only continue once that input is received.
This matters because a large share of real business processes are not fully automatable and should not be. Client onboarding, refunds above a threshold, content that goes out under a brand's name, hiring steps — all benefit from an explicit human checkpoint. Relay lets teams build these as durable, documented workflows where the automated and human steps are interleaved on purpose. Reviewers repeatedly single this out as the reason they chose Relay over alternatives, and in our assessment it is a genuine, defensible differentiator rather than a marketing line.
The AI Builder and AI Steps
Relay leans into AI in two distinct ways. First, an AI builder lets you describe the automation you want in plain language; the assistant, which runs on current frontier models, constructs an editable visual workflow, prompts you to connect the right apps, and proactively flags issues. For non-technical users this dramatically lowers the barrier to a working automation — you are refining a draft rather than assembling nodes from scratch.
Second, AI steps let a workflow call a model mid-run to draft, summarise, classify or transform data. These are powered by included AI credits that grant access to models from OpenAI, Anthropic and Google without the user having to set up separate provider accounts. If you already have an API key, you can bring it. The combination — AI to build the workflow and AI inside the workflow — is what earns the platform its "AI-native" description, and it is more than skin deep.
Pricing: Transparent and Aggressive
Relay's pricing is one of its strongest cards. There is a genuine free plan — 200 steps and 500 AI credits a month with access to every integration. On annual billing, Professional is $19 a month and Team is $59 a month for up to ten users, with Relay advertising a 50% saving versus monthly billing. Enterprise is custom.
Two design choices make this more generous than the headline numbers suggest. First, every integration is available on every plan, including Free — there is no premium-connector upcharge, which is a common frustration with legacy tools. Second, Relay does not count triggers, data transformations, or path and loop evaluations against your step budget, so a workflow with many logical branches consumes far fewer steps than a naive count would imply. For small and mid-sized teams, the effective cost of running real automations on Relay is markedly lower than on the incumbents.
Integrations and Extensibility
Relay offers more than 200 app connectors covering the tools most teams use daily — Gmail, Slack, Notion, HubSpot, Salesforce, Airtable, Linear and many more — alongside generic HTTP requests, webhooks and Model Context Protocol support for connecting custom tools. For technical users there are custom JavaScript steps, a secret store, and rich data-transformation and file utilities.
The honest caveat is breadth. Zapier has spent more than a decade building the largest connector library in the category, and Relay, as a younger platform, does not match it on sheer count. For the long tail of niche SaaS apps, a team may occasionally find a connector missing. Relay's generic HTTP and webhook support covers many of those gaps, and reviewers note the team ships requested integrations unusually fast, but buyers with exotic tool stacks should confirm coverage before committing.
Security, Compliance and Enterprise Readiness
Relay.app states that it is SOC 2 compliant and that customer data is not used to train AI models. SSO for Microsoft and Google, custom run-history retention, and user management arrive on higher tiers, and Enterprise adds GDPR compliance, custom usage limits, custom integrations and onboarding workshops. This is a reasonable enterprise posture for a platform of Relay's maturity, though buyers who need a signed SOC 2 report and SSO on an entry-level plan will need to move to Enterprise sooner than they might on some larger vendors.
Who Relay.app Is Really For
Relay.app is at its best for operations, revenue and support teams that need automations combining machine speed with human judgement, and that value transparent, affordable pricing. It is especially attractive to teams migrating away from a legacy tool that felt expensive or clumsy. It is a weaker fit for organisations that depend on a very long tail of niche connectors, or that run automations at a volume that would repeatedly blow through standard step limits. For most small and mid-sized teams evaluating a modern automation platform in 2026, Relay belongs on the shortlist — and its free tier makes it trivial to prove out before spending anything.
