An AI agent that tests applications or APIs using plain English goals and generates Playwright test specifications upon success.
Also known as sekharsdet/five46
- Stars
- 3
- Tools
- 2
- Downloads, 30 days
- —
- Latest version
- 0.2.3
- Last commit
- Oct 19, 2026
- License
- MIT
What is five46 MCP?
five46 is an AI agent that tests applications or APIs using plain English goals and generates Playwright test specifications upon success. It exposes 2 tools, such as five46_test, five46_api.
Summary generated from the project's README
About five46 MCP server
Automating application and API testing through plain English goals is what five46 enables for developers and QA engineers. By providing a natural language objective, such as logging into a dashboard or verifying a POST request, the agent uses your own LLM keys from providers like OpenAI, Anthropic, Gemini, Groq, or AWS Bedrock to drive a real browser or interact with an API. It operates locally via a CLI or through an MCP server for IDE assistants, ensuring your application data only reaches your chosen LLM provider.
When a goal is successfully met, the tool automatically generates a standalone Playwright or node:test specification file. This allows you to convert a one-time manual check into a permanent regression test that runs without needing an LLM or the five46 agent. The system provides a pass or fail verdict, screenshots, and a root-cause hypothesis for failures. For API testing, the agent remains read-only by default unless you explicitly use the allow-writes flag.
What you can do with it
- 1Convert natural language goals to Playwright testsA developer can describe a user flow in English and receive a ready-to-commit test specification file.
- 2Verify API functionality with plain EnglishYou can instruct the agent to perform specific API actions like creating a user and then verifying the creation via a GET request.
- 3Debug failed application flowsWhen a test fails, the agent provides a hypothesis regarding the underlying cause to help speed up troubleshooting.
Written from the project's README. Every use case is checked against a quote from it.
How to install five46 MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall five46 MCP in Claude Code
$ claude mcp add --env 'FIVE46_LLM_API_KEY=<YOUR_FIVE46_LLM_API_KEY>' --env 'FIVE46_LLM_PROVIDER=<FIVE46_LLM_PROVIDER>' five46 -- npx -y five46 mcpInstall five46 MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"five46": {
"command": "npx",
"args": [
"-y",
"five46",
"mcp"
],
"env": {
"FIVE46_LLM_API_KEY": "<YOUR_FIVE46_LLM_API_KEY>",
"FIVE46_LLM_PROVIDER": "<FIVE46_LLM_PROVIDER>"
}
}
}
}Install five46 MCP in VS Code
.vscode/mcp.json
{
"servers": {
"five46": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"five46",
"mcp"
],
"env": {
"FIVE46_LLM_API_KEY": "<YOUR_FIVE46_LLM_API_KEY>",
"FIVE46_LLM_PROVIDER": "<FIVE46_LLM_PROVIDER>"
}
}
}
}Install five46 MCP in VS Code CLI
$ code --add-mcp '{"name":"five46","type":"stdio","command":"npx","args":["-y","five46","mcp"],"env":{"FIVE46_LLM_API_KEY":"<YOUR_FIVE46_LLM_API_KEY>","FIVE46_LLM_PROVIDER":"<FIVE46_LLM_PROVIDER>"}}'Install five46 MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"five46": {
"command": "npx",
"args": [
"-y",
"five46",
"mcp"
],
"env": {
"FIVE46_LLM_API_KEY": "<YOUR_FIVE46_LLM_API_KEY>",
"FIVE46_LLM_PROVIDER": "<FIVE46_LLM_PROVIDER>"
}
}
}
}Install five46 MCP in Codex
$ codex mcp add five46 --env 'FIVE46_LLM_API_KEY=<YOUR_FIVE46_LLM_API_KEY>' --env 'FIVE46_LLM_PROVIDER=<FIVE46_LLM_PROVIDER>' -- npx -y five46 mcpInstall five46 MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.five46] command = "npx" args = ["-y", "five46", "mcp"] [mcp_servers.five46.env] FIVE46_LLM_API_KEY = "<YOUR_FIVE46_LLM_API_KEY>" FIVE46_LLM_PROVIDER = "<FIVE46_LLM_PROVIDER>"
Install five46 MCP in Gemini CLI
$ gemini mcp add -s user -e 'FIVE46_LLM_API_KEY=<YOUR_FIVE46_LLM_API_KEY>' -e 'FIVE46_LLM_PROVIDER=<FIVE46_LLM_PROVIDER>' five46 npx -- -y five46 mcpInstall five46 MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"five46": {
"command": "npx",
"args": [
"-y",
"five46",
"mcp"
],
"env": {
"FIVE46_LLM_API_KEY": "<YOUR_FIVE46_LLM_API_KEY>",
"FIVE46_LLM_PROVIDER": "<FIVE46_LLM_PROVIDER>"
}
}
}
}Started by HostedMCP on 16 Nov 2026
We ran this command for five46 0.2.3 in an isolated container with no network and no keys. It answered and listed 2 tools.
five46 MCP tools 2
Read from the server with tools/listfive46_testDrive a real local Playwright browser toward a stated goal, one action at a time, using your own configured LLM key (BYOK). Writes a real, standalone, re-runnable Playwright spec on success. Discloses: the page's visible text/labels/values are sent to your configured LLM provider on every step. Clicking an element that looks like it deletes/deactivates/closes an account or permanently erases data (e.g. "Delete Account") is blocked for this server — set FIVE46_MCP_ALLOW_DELETES=1 in the server environment to unlock it. Never settable via a tool call argument. This is a best-effort heuristic on the clicked element's visible text, not a guarantee. On a real assertion failure (a genuine finding about the app, not a tooling issue), one additional LLM call analyzes the failure for a root-cause hypothesis — built only from information already visible to the agent during the run (never a new disclosure), always on for this tool, and included in the returned report as a hypothesis, not a confirmed diagnosis.
five46_apiDrive real HTTP requests toward a stated goal, one request/assertion at a time, using your own configured LLM key (BYOK). Writes a real, standalone node:test script on success. Read-only for this server (GET/HEAD/OPTIONS only) — set FIVE46_MCP_ALLOW_WRITES=1 in the server environment to unlock POST/PUT/PATCH, FIVE46_MCP_ALLOW_DELETES=1 to separately unlock DELETE. Neither is ever settable via a tool call argument. Requests are restricted to the target base URL's own origin — no additional host allowlist is available via this tool. Discloses: request/response data, including response bodies (which may contain a live session token or other secret), is sent to your configured LLM provider on every step. On a real assertion failure, one additional LLM call analyzes the failure for a root-cause hypothesis — built only from information already visible to the agent during the run, always on for this tool, and included in the returned report as a hypothesis, not a confirmed diagnosis.
Configuration
FIVE46_LLM_API_KEYRequiredSecretYour own API key for the chosen provider (for bedrock, your AWS region instead)
FIVE46_LLM_PROVIDERRequiredLLM provider to use: openai, anthropic, gemini, groq, or bedrock
FIVE46_MCP_ALLOW_DELETESSet to 1 to unlock DELETE (five46_api) and destructive-looking browser clicks (five46_test)
FIVE46_MCP_ALLOW_WRITESSet to 1 to unlock POST/PUT/PATCH for the five46_api tool (read-only by default)
FIVE46_MCP_CONCURRENCYMax concurrent scenarios for story mode (default: 2 for five46_test, 3 for five46_api; hard-capped at 5)
Automated checks
Ran Nov 13, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachablesekharsdet/five46 is public and reachablegithub.com/sekharsdet/five46 ↗Pass
- Package names this repository as its sourcefive46 names sekharsdet/five46 as sourcenpmjs.com/package/five46 ↗Pass
- Build provenance (Sigstore attestation)No build attestation published for this releaseNot checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-19github.com/sekharsdet/five46 ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNo popular package one edit away from five46Pass
- No critical advisory in the package or its dependencies129 resolved dependencies checked; no critical advisoryosv.dev ↗Pass
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubfive46
Say what should work, in plain English. An AI agent checks it against your real app or API, using your own LLM key, and hands you a real Playwright test.

You just built something and want to know whether it actually works, before
writing a test. Give five46 a goal like "log in and confirm the dashboard
loads" or "create a user via POST, then confirm it via GET". An LLM
(OpenAI, Anthropic, Gemini, Groq or AWS Bedrock, with your key) drives your
real app or API one action at a time. It tells you honestly whether the goal
was met, with a root-cause hypothesis when it wasn't. A successful run is
saved as a standalone Playwright (or node:test) spec, so today's check
becomes tomorrow's regression test, and running it needs no five46 or LLM.
Quick start
npm install -g five46
npx playwright install chromium # one-time browser download
five46 config # pick a provider and paste a key (Gemini's is free)
five46 doctor # confirms the key works
five46 test http://localhost:3000 --goal "log in as demo@example.com and confirm the dashboard says Welcome back"
You get a pass or fail (and a matching exit code), a report with
screenshots and, for failures, a root-cause hypothesis, plus a
five46-agent-*.spec.ts you can commit and run with npx playwright test.
Testing an API instead? Use five46 api <base-url> --goal "...". It's
read-only unless you pass --allow-writes.
Status: early and actively developed. Verified end-to-end against dozens of real-world sites and APIs.
Check every pull request
The five46 GitHub Action reads the Acceptance criteria section of a pull request's description and checks each item against the PR's preview deployment. It keeps one PR comment up to date with the verdict per criterion, why each failure failed, and a ready-to-commit spec for every criterion that passed.
on: deployment_status # Vercel, Netlify and other preview deploys
jobs:
five46:
if: github.event.deployment_status.state == 'success'
runs-on: ubuntu-latest
permissions: { contents: read, pull-requests: write }
steps:
- uses: sekharsdet/five46-action@v1
with:
url: ${{ github.event.deployment_status.target_url }}
api-key: ${{ secrets.FIVE46_API_KEY }}
How it compares
| five46 | Playwright Test Agents | playwright codegen |
Cloud AI testing platforms | |
|---|---|---|---|---|
| You provide | A goal or acceptance criteria in plain English | A seed test and a request, inside an AI coding tool | Your own clicks | Test descriptions in their app |
| Runs | CLI, CI, GitHub Action or MCP tool | Inside an AI coding tool, such as VS Code | Locally, while you click | Their cloud |
| Built to give a pass/fail verdict on the app | Yes, with a root-cause hypothesis | Not its focus: it generates and repairs tests | No | Yes |
| API testing | Yes (node:test output) |
Browser-focused | No | Varies |
| Proves assertions catch failures | Yes (--verify mutation check) |
No built-in check | No | Varies |
| LLM and cost | Your key, any of 5 providers | Your coding agent's model | None | Their subscription |
| Your app data goes to | Only your LLM provider | Your coding agent's provider | Nowhere | Their platform |
| Output | Plain Playwright or node:test files |
Plain Playwright files | Plain Playwright files | Usually their runner |
Already working in an AI coding tool and growing a Playwright suite there? Playwright's own agents are a great fit, especially the healer for keeping existing tests green. five46 is for getting a verdict: is this feature working right now, in CI or on a PR, for a browser flow or an API, with any LLM you already pay for. More detail is in Why five46.
Highlights
- Safe by default. API runs are read-only unless you unlock writes; destructive-looking clicks ("Delete account") are blocked unless allowed.
- Login credentials never reach the LLM. It sees placeholders, and the real values are filled in locally. Log in once, then reuse the session.
- Acceptance criteria in, verdicts out.
--storysplits a user story or ticket into independent scenarios and reports pass/fail per criterion. - Specs you can trust.
--verify-clean-sessionre-runs the spec cold;--verifymutates its assertions and requires them to fail. - Works with any current key. If your provider retires a model, five46
finds a current one your key can use. You can also pin one with
FIVE46_LLM_MODEL. - Fits your tools. CI-friendly exit codes,
--reportJSON for dashboards, flaky-run detection (--repeat), video recording, and an MCP server for IDE assistants.
Docs
- Configuration: providers, getting a key,
five46 doctor, choosing models - CLI reference: every command and flag, login sessions, API testing, story mode, verification, reports
- MCP server: use five46 from an IDE-embedded AI assistant
- Security model: what leaves your machine, what five46 guards, and what's your call
- Why five46: design choices, the full feature list, comparisons
- CHANGELOG
Known limitations
- No closed shadow DOM traversal. Elements inside a
mode: "closed"shadow root aren't visible to the agent's page snapshot (a closed shadow root is deliberately inaccessible to external tooling in general, not just five46). Open shadow roots, the overwhelming majority of real-world web components, work fine, as do<iframe>and<frame>content. - Chromium only for browser mode, with no Firefox or WebKit yet.
- A run itself is not deterministic. The same goal against the same page can take a different path next time. The generated spec is the frozen, repeatable artifact.
Contributing
Issues and pull requests are welcome. See CONTRIBUTING.md.
git clone https://github.com/sekharsdet/five46.git
cd five46
npm install
npm run build
node dist/cli.js test <url> --goal "..."
npm test
If five46 is useful to you, a ⭐ on GitHub helps other people find it.
License
five46 MCP questions
How do I add five46 MCP to Claude Code?
Run claude mcp add --env 'FIVE46_LLM_API_KEY=<YOUR_FIVE46_LLM_API_KEY>' --env 'FIVE46_LLM_PROVIDER=<FIVE46_LLM_PROVIDER>' five46 -- npx -y five46 mcp in your terminal. Add --scope user to make it available in every project.
How do I add five46 MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What tools does five46 MCP provide?
It exposes 2 tools, including five46_test, five46_api.
How does five46 handle API testing?
You can test an API using the command five46 api <base-url> --goal "...", which is read-only by default unless you explicitly pass the --allow-writes flag.
What happens if the test fails?
The tool provides a pass or fail verdict, an exit code, a report with screenshots, and a root-cause hypothesis for the failure.
Does it work with my own LLM keys?
Yes, it uses your own LLM key from providers like OpenAI, Anthropic, Gemini, Groq, or AWS Bedrock to drive the application.
Can I use the generated tests without five46?
Yes, successful runs are saved as standalone Playwright or node:test specifications that can be run independently.