Model Context Protocol server
UX testing inside Claude Code, Cursor, and Codex
Swarm’s MCP server gives your coding agent a real testing loop. AI personas navigate your localhost, find friction, and send structured UX issues back into the chat. The model that wrote the code is the one that sees it break.
Included on Startup ($150/mo). Not on the Free plan.
- Setup
- 3 min
- First results
- ~10 min
- Live runs on Startup
- 20 / mo
Streaming findings from 5 personas…
Email accepts obvious typos like foo@gmial.com
High · Step 1 of 5
Password rules hidden until error
Medium · Step 2 of 5
Loading state delays 800ms after submit
Medium · Step 4 of 5
One command, every editor
The Swarm MCP server is published to npm as @useswarm/mcp. The setup command registers Swarm with every MCP-capable editor it detects and signs you in through the browser.
- Runs locally over stdio
- Credentials stored at ~/.useswarm/config.json
- No source code leaves your machine
Auto-configured by the setup command. Restart the editor and you are done.
$ npx @useswarm/mcp@latest setupDetects Claude Code and Codex CLI automatically. Add --client claude to scope to one editor.
From prompt to findings, without leaving chat
The MCP server runs locally over stdio. It tunnels your localhost through Cloudflare so Swarm's cloud agents can reach your dev server. No code leaves your machine.
- 1.
You prompt your editor
In Claude Code or Cursor, you write a one-line request. The model recognizes it as a Swarm tool call and invokes dev_test with a target URL, a goal, and an audience.
- 2.
The MCP tunnels your localhost
The local MCP server opens a short-lived Cloudflare tunnel so Swarm's cloud-hosted agents can reach your dev server. The tunnel closes when the test ends.
- 3.
AI personas navigate your app
Personas matching the audience you described navigate the flow. They click, type, scroll, and read the page like real users, surfacing confusing copy, broken interactions, and drop-off points.
- 4.
Structured issues stream back
Your editor receives a stream of UX findings as they happen. Each finding has a severity, a description, the step it occurred at, and a specific suggested fix.
- 5.
Fix, then re-test
When the model edits code and the dev server reloads, you can re-run the same test with one line. Same audience, same goal, see whether the issues are gone.
The tools your editor gets
Once the MCP is connected, the model calls these as part of its normal tool use. You don't have to memorize them. Just ask the editor to test something.
dev_test
(url, goal, audience?) → run_id
Start a test run. Takes a target URL, a goal in plain English, and an optional audience description.
dev_watch
(run_id) → stream<finding>
Stream structured UX issues as the personas find them. Each finding includes a severity, location, and suggested fix.
dev_status
(run_id) → progress
Check progress on an in-flight test. How many personas have run, how many issues found so far.
dev_list_swarms
() → run[]
List the recent test runs for this project. The model uses this to pick up where it left off across sessions.
dev_close
(run_id) → ok
End an active session and tear down the Cloudflare tunnel. Called automatically when a test completes.
What teams use it for
Prompt
Test the signup flow you just shipped. Audience: first-time SaaS users.
Test the flow you just wrote
After Claude or Cursor writes a new signup, checkout, or onboarding flow, ask it to test what it just shipped before you reload the page yourself.
Prompt
Same test as last time. Did anything break?
Catch regressions while iterating
When you refactor, the model reruns the last test against your dev server and tells you whether anything broke, without a full QA cycle.
Prompt
Verify the new CTA copy actually lands with the target audience.
Validate AI-generated UI
When the model proposes a UI change, it can verify that real personas in the target audience actually understand and click through.
Questions, answered
Setup, security, supported editors, and what each plan includes. Something else? Email the founders
An MCP server is a process that speaks the Model Context Protocol, an open standard from Anthropic that lets AI tools like Claude Code, Cursor, and Codex call external tools and read external data. Swarm's MCP server exposes UX testing tools so your AI coding assistant can run real usability checks against your app without leaving the editor.
Run npx @useswarm/mcp@latest setup. The one command registers Swarm with Claude Code and Codex CLI and signs you in. After it finishes, restart your editor and ask it to test your app.
Claude Code, Cursor, Codex CLI, and any editor that supports the Model Context Protocol. The setup command auto-configures Claude Code and Codex. Manual config snippets are provided for everything else above.
No. The MCP server only needs the URL of your running dev server (typically localhost:3000). It tunnels your localhost via Cloudflare so Swarm's cloud-hosted AI agents can navigate the app. Your source code never leaves your machine.
The main tools are dev_test (start a test run with a goal, audience, and target URL), dev_watch (stream structured UX issues as the agents find them), dev_status (check progress), dev_list_swarms (browse prior runs), and dev_close (end a session). Your editor's AI calls them as part of normal tool use.
Yes. MCP server access is included on the Startup plan ($150/month). The Free tier does not include MCP access. Startup gives you 20 live agent runs per month, which is what the MCP integration uses, plus 50 screenshot runs.
Ship code, then test it from the same chat
MCP access is included on Startup ($150/mo), with 20 live agent runs and 50 screenshot runs a month. Enough to wire the loop into your editor and keep testing as you ship.
- Startup plan
- $150/mo
- Live runs a month
- 20
- First results
- ~10 min