Relay Get early access
Rely on Relay · local-first API & UI testing

Ship the product your spec promised.

Relay reads your requirements, checks every sentence of them on each run, and learns your app as it goes. Your API calls, browser flows and the spec that governs them — one tool, all in your repo.

Get early access See it work

Free to start. Works offline. Collections and tests are plain files you commit, not a cloud account.

relay · product health
Your data never leaves your machine Everything versioned in Git Runs offline and in CI Open, inspectable, no lock-in

See your whole system

Relay maps every screen, API and the links between them.

As your requests and flows run, Relay builds a living graph of your app — and measures your spec against it. Here it is on the example project that ships with Relay:

172 / 185

requirement sentences checked on every run — the other 13 are flagged, never silently skipped. (Figures from the bundled sample project.)

Sign in/signin Dashboard/home Orders/orders Customers/customers Checkout/checkout Order detail/orders/:id API/api/auth API/api/orders
Screen, tested API behind it live signal

What makes Relay different

Most tools check that an endpoint replied. Relay checks that your product did what the spec said.

Requirements, enforced

Every sentence of your spec, checked on every run.

Point Relay at your requirements document. It turns each sentence into a numbered statement and shows you exactly which ones are covered, which are not, and which are only tested from one side — where a bug would slip through.

  • A live coverage score for the whole document
  • AI writes the missing checks — your own key, or a local model, free
  • Kept only when they hold on a build you trust
Relay's Requirements screen: each statement of the spec marked checked or not, with a coverage score and an AI model picker.

It learns your app

A living map of every screen, action and API.

As your tests run, Relay builds a model of how your app really fits together — the pages, what connects to what, the APIs behind them. Change one endpoint and it tells you exactly which tests and screens are affected before you ship.

  • See what a change affects, as a pull-request summary
  • Catches contract drift and screens that disagree with their data
  • Learns from real production traffic, too
Relay's app map: a graph of the application's screens and the connections between them, with coverage stats.

One place, one keystroke

The whole product, at a glance — and anything you need, instantly.

An overview that opens on what matters: what your spec keeps, what still needs you, what Relay has learned. And a command bar that reaches every request, flow and action without hunting through menus.

  • Imports your existing collections and OpenAPI specs
  • Browser flows that repair their own broken selectors
  • Finds the bugs other tools miss: data-grid sorting and filters, missing validation styling, interface lag
Relay's command bar: search requests, flows and requirements, or run any action from one keystroke.

Proof

Catch the bugs your other tools can’t see.

Point Relay at a page and it exercises what actually breaks in real apps — data grids, forms, interaction speed — then tells you exactly what is wrong, in plain language. One command, real output:

Relay ui check output: four high-severity findings — a grid sorting numbers as text, a filter hiding half the matching rows, a required field with no red border, and a 1.4-second click freeze.
Collections and tests as plain files in your repo
Works fully offline, no account to sign into
Requirements turned into checks, with live coverage
Learns your app as a living map
API testing, mock servers and load tests
Browser flows that repair their own selectors
AI that writes checks — your key, or a local model
Callable by AI agents over MCP and UTCP
Data-grid, form-validation and lag checks
Signed evidence documents and CI reporting

Built for the agent era

Give your coding agent a testing toolkit.

Relay speaks MCP and the newer UTCP, so GitHub Copilot, Cursor, Claude or your own agent can drive it — safely, inside your workspace. One command connects them, and they get real tools, not guesses.

GitHub Copilot Cursor Claude MCP UTCP
› Check which requirements in requirements.md aren’t tested yet.
› Probe localhost:3000 and tell me what breaks.
› Run the orders collection and summarise the failures.
› What does changing POST /api/orders affect?
# connect Relay to your agent
$ relay mcp setup ./my-project
  ✓ GitHub Copilot (VS Code)
  ✓ Cursor  ✓ Claude  ✓ Cursor CLI

# or expose it over HTTP (UTCP)
$ relay utcp ./my-project
  manual → 127.0.0.1:4477/utcp

send_request

Fire any HTTP request, with your environment variables filled in.

run_collection

Run a whole collection or folder with its tests, like a CI pass.

requirements_coverage

Report which spec sentences are checked, and which are not.

check_page

Open a page and check its grids, forms and interaction speed.

probe_api

Run the decision tables and rules against a live build.

change_impact

Show what a change to an API or screen affects downstream.

Get started

Up and running in two minutes.

Relay is in early access. Here is how it installs — join the list and we will send you the build.

Desktop app

macOS Windows Linux

Collections and tests land as plain files in your repo.

Command line & CI

# install the runner $ npm i -g relayqa # set up a project $ relay init ./my-project # run it anywhere $ relay req coverage requirements.md $ relay flow run flows/ --since origin/main

Early access

Be first to rely on Relay.

Relay is in early access. Leave your email and we’ll send you the build and the founder price before anyone else.

Pricing

Start free. Own your copy. No seats to count.

Relay runs on your machine and your CI. Launch pricing below — early adopters keep it.

Free

For trying Relay and personal projects.

$0
  • API & UI testing, one project
  • Requirements coverage
  • Local AI model support
  • Community support
Get early access

Pro

For working engineers who want the whole toolkit.

$129 / one-time, per seat
  • Everything in Free, yours to keep
  • Unlimited projects, local and in CI
  • One year of updates included
  • AI check generation with any provider
  • Evidence documents & CI reporting
  • Agent access (MCP / UTCP)
Request Pro access

Team

For teams who ship together and need the audit trail.

$3,000 / year · unlimited seats
  • Everything in Pro
  • Shared design systems & templates
  • Priority support
Talk to us

Documentation

Everything is written down, and worth reading.

From your first request to wiring Relay into CI and agents — clear, example-led guides. Here is where most people start.

01

Quickstart

Open a project, send your first request, and run a flow in five minutes.
02

Requirements & coverage

Turn a spec into checks and see what your product keeps, line by line.
03

The app map

How Relay learns your screens, APIs and what a change will affect.
04

Grid, form & lag checks

Catch the interface bugs that unit tests and API checks never see.
05

Agents: MCP & UTCP

Connect Copilot, Cursor or Claude and let them drive Relay safely.
06

CLI & CI

Every command, and how to run Relay on each pull request.

Read the docs

Rely on Relay.

Point it at a project and watch it tell you — line by line — what your product actually delivers. Your spec is a promise; keep it.

Get early access See it work