# Usability testing MCP > An MCP server that lets your coding agent order a usability test with a real person and read the session back as evidence it can act on. Page: https://hallwaytest.ai/usability-testing-mcp MCP endpoint: https://mcp.hallwaytest.ai/mcp Last updated: November 9, 2026 **Short answer:** A usability testing MCP is a Model Context Protocol server that gives a coding agent tools to run usability tests with real people. With Hallway Test, Claude Code, Cursor or Codex sends a URL and a task to an in-house tester, then pulls the recording, transcript and event log back within 24 hours and turns them into fixes. ## How it works 1. **Your agent plans the test.** `plan_usability_test_suite` has the agent read your project and draft 5 to 10 critical user flows, each written as a situation a stranger can follow, with a clear finish line. 2. **A real person runs it.** `create_usability_task` sends one flow to an in-house tester. They use your product for 15 to 30 minutes while their screen, voice and every click are recorded. Hallway Test receives a URL (or a mobile app key) and the task, never your source code. 3. **The session comes back to the agent.** The recording is ready within 24 hours, and `get_usability_result` brings it into the agent whenever you ask for the status. The agent then pulls the evidence with the Jam tools for web apps or the `mobile_*` tools for React Native and Expo apps. 4. **The agent proposes fixes, you choose.** `plan_jam_ux_audit` makes the agent prove each friction point from the event log and transcript, link it to the exact second of the video, and rank the fixes. Nothing in your code changes until you pick which fixes to make. ## What comes back - A screen recording of the session, with the tester’s webcam and voice - A timestamped transcript of what the tester said while thinking aloud - A full log of clicks, taps, typing and navigation, plus console and network logs - In your agent: a short summary of how hard the task was, and a table of fixes with linked timecodes and a severity (conversion risk, friction or polish) The audit follows a strict order: events first, then the transcript, then screenshots, and console or network logs only to explain a problem the events already show. A friction point that the evidence does not support is left out. ## All 22 tools ### Run a test - `create_usability_task`: Writes and dispatches a real usability test to a real user, based on the flow you describe. - `get_usability_result`: Checks a test right away: still with the tester, failed, or the recording ready to audit. - `list_usability_tasks`: Lists your recent tests with their status, so you can ask how they are going without a task ID. - `wait_for_usability_result`: For clients that keep long requests open: follows the run and hands back the recording as soon as it’s ready. - `cancel_test_run`: Cancels a test that was already dispatched. ### Plan tests - `plan_usability_test_suite`: Scans your project and drafts a full test suite – 5 to 10 critical user flows, ranked by risk. ### Analyze a recording - `plan_jam_ux_audit`: Turns a raw recording into a structured, evidence-backed UX audit by running the tools below in the right order. - `jam_get_details`: Recording metadata – type, duration, whether a voice transcript exists, and the shareable link. - `jam_get_user_events`: The full click / type / navigation log – the source of truth for what the tester actually did. - `jam_get_video_transcript`: The tester’s spoken think-aloud commentary, timestamped. - `jam_get_frames`: Pulls screenshots from key moments in the recording to visually confirm a friction point. - `jam_get_console_logs`: Browser console logs, used to explain a technical cause behind friction that’s already been proven. - `jam_get_network_requests`: Network request log, used to trace a failed call behind an already-evidenced UX problem. ### Mobile apps - `setup_mobile_sdk`: Adds the Hallway Test SDK to your React Native or Expo app, with your app key filled in, and confirms the first recording arrived. - `list_mobile_apps`: Your mobile apps and their app keys, including apps a teammate shared with you. - `mobile_get_details`: Start of every mobile audit: device, app version, duration, crash, what was recorded, and the viewer link. - `mobile_get_user_events`: Taps, swipes, text input, screens and navigation, each at its second in the video – the source of truth for a mobile audit. - `mobile_get_video_transcript`: What the person said while recording, timed to the video. - `mobile_get_frames`: Screenshots from the screen video at the seconds the events point to. - `mobile_get_errors`: JavaScript errors, native crashes and frozen screens, each with its time in the video. - `mobile_get_console_logs`: JavaScript console output and native logs, filtered by level. - `mobile_get_network_requests`: HTTP and WebSocket traffic of the app, used to explain a spinner that never ends or a submit that does nothing. ## Connect it Claude Code: `claude mcp add --transport http hallwaytest https://mcp.hallwaytest.ai/mcp` - [Connect Hallway Test to Claude](https://hallwaytest.ai/connect-mcp.md) - [Connect Hallway Test to Cursor](https://hallwaytest.ai/connect-cursor.md) - [Connect Hallway Test to Codex](https://hallwaytest.ai/connect-codex.md) A new account can connect right away, but the first request to dispatch a test answers with a link to book the free pilot call. After the 30-minute call we activate the account, with no need to reconnect: you can send tests right away, and results come back within 24 hours. ## What it costs Free pilot: 1 usability test + 30-minute feedback call. Planned price after public launch: $29 per test, no subscription. The analysis runs in your own agent, so Hallway Test does not charge for it. ## FAQ **What is MCP?** MCP, the Model Context Protocol, is an open standard for connecting AI applications to outside tools and data. An MCP server describes its tools, and any compatible client, such as Claude Code, Claude, Cursor, Codex or VS Code, can call them. Hallway Test is a remote MCP server: you add one URL and sign in with Google. **Is the tester a real person or an AI?** A real person. Every test is run by an in-house tester who has never seen your product. AI only enters on your side, when your own agent reads the session and writes up the friction points. **Who analyzes the session, and who pays for that?** Your coding agent, on your own subscription or API key. The Hallway Test server runs no AI model: its tools return the recording data plus instructions on how to audit it, and your agent does the analysis. **Does Hallway Test see my code?** No. Planning happens inside your agent. To order a test, the agent sends only the address of the build to test (or a mobile app key) and the task text, plus test login details if the flow needs them. **Can it test mobile apps?** Yes, React Native and Expo apps on iOS and Android. The setup_mobile_sdk tool has your agent add the Hallway Test SDK, which records the screen, voice, taps, screens, network calls and crashes during the session. **How long does a test take?** The tester spends 15 to 30 minutes on one flow, and the results come back within 24 hours. You can keep working in the meantime and ask your agent for the status at any time. ## Related - [Home](https://hallwaytest.ai/index.md) - [Pricing](https://hallwaytest.ai/pricing.md) - [Book a free pilot](https://hallwaytest.ai/book.md) - [UX audit with a real user](https://hallwaytest.ai/ux-audit.md) - [Website usability testing service](https://hallwaytest.ai/website-usability-testing.md) - [All guides and comparisons](https://hallwaytest.ai/learn.md)