# Functionize MCP server (slim)

> The Functionize MCP server is a hosted remote server from Functionize, Inc. that lets coding agents such as Claude Code, Cursor and GitHub Copilot build, run and diagnose browser tests of a live web app in Functionize Studio.

- Full: https://www.anchorterminal.com/tools/functionize-mcp.md (~11,200 tokens) · this version ~1,880 tokens · JSON https://www.anchorterminal.com/tools/functionize-mcp.json · canonical https://www.anchorterminal.com/tools/functionize-mcp
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-25 · pricing https://www.anchorterminal.com/tools/functionize-mcp#pricing · reviews https://www.anchorterminal.com/tools/functionize-mcp#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**E · 44.7/100 · rank #1031 of 1210 · #18 in Browser automation · not agent-ready · confidence low**


Assessment: The server passes a coding agent's test requests to Functionize Studio's testing agent, which tests a live web app in cloud browsers and reports the failing step. It uses OAuth with read and write scopes and has a no-card free plan. Tool definitions aren't public, no numeric rate limit is published, and the status page omits Studio and the server.

## Facts

- Kind: MCP server · vendor: Functionize, Inc. · category: Browser automation · legal entity: Functionize, Inc. · provenance 80/100
- Endpoint: `https://mcp.functionize.com/mcp` (Streamable HTTP)
- Auth: OAuth · pricing: Freemium · x402: no · licence: Proprietary hosted service under Functionize's Terms of Service (version 2.0, effective 10 July 2026). The Claude Code plugin and its two skills are MIT-licensed; the server's source isn't public
- Probe metrics: not measured yet (probes haven't run)
- MCP endpoint: `https://mcp.functionize.com/mcp`, Streamable HTTP. Unauthenticated `initialize` and `tools/list` answer 401 with resource metadata. `/health` answers `ok`
- Tools: 10 per the plugin README. Named in the skills: `start_agent_session`, `send_agent_message`, `get_agent_session`, `get_agent_session_events`, `stream_agent_session_events`, `stop_agent_session`, `list_agent_sessions`, `list_agent_teams`, `upload_session_file`, `delete_session_file`
- OAuth: Authorisation code with PKCE S256, refresh tokens, dynamic client registration, client ID metadata documents, public clients, revocation endpoint. Scopes `agent.sessions:read`, `agent.sessions:write`, `offline_access`
- Team scope: Account default team, or the team in `X-Functionize-Team-Id`, membership checked per request. Sessions are visible and controllable across the team
- Claude Code plugin: `functionize-mcp` 1.1.0, MIT, from the FunctionizeInc/functionize-mcp-plugin marketplace. Bundles the connection and two skills, `functionize` 2.12.0 and `functionize-prompting` 2.40.0
- Clients documented: Claude Code, Claude Desktop, claude.ai, Cursor, GitHub Copilot in VS Code and Visual Studio 2022, Gemini CLI, Devin, Codex CLI and Windsurf configs; Cursor web and Cloud Agents enabled but unconfirmed
- Browsers: Chrome, Firefox and Edge on cloud-managed browsers, per the plugin's capabilities reference. Web UIs only
- Files: Up to 5 MiB per file, sent by the client; the server has no file-system access
- Status: status.functionize.com on incident.io, components app.functionize.com, app.virtualautomationengineer.com and uk.virtualautomationengineer.com
- Prices: Studio Growth plan $20 per month (plan); Studio Scale plan $100 per month (plan); Add-on credits, Growth and Scale $0.0001 per credit

- Scores: Reliability 50, Performance pending, Schema & documentation 29, Agent ergonomics 43, Security & auth 56, Payments & pricing 30, Task success pending, Maintenance & community 49, Transparency & trust 59 · total over the 7 assessed categories
- Why: Reliability, Graded with the hosted lines, on the remote MCP server. · Schema & documentation, No machine-readable contract could be read. · Agent ergonomics, Graded on the MCP server. · Security & auth, OAuth only. · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, Read as a closed hosted service with a public plugin repository. · Transparency & trust, The editorial half.
- Sources: 18, open questions: 6, both in the full twin
- Capabilities: browser.control, browser.hosted, browser.debug
- JSON: https://www.anchorterminal.com/api/v1/tools/functionize-mcp.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/functionize-mcp.svg` or a link to https://www.anchorterminal.com/tools/functionize-mcp from a page on functionize.com or one of its subdomains, or the README of github.com/FunctionizeInc/functionize-mcp-plugin, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Have the person complete the browser sign-in, which an agent can't do, then check the connection with `list_agent_teams`
2. Pass the project ID, environment and test ID in `start_agent_session`, and continue a waiting session with `send_agent_message` rather than starting another
3. Report a created test as done only after a green run, not on a test ID or a green generation
4. After a dropped connection, list the project's recent tests before creating a test again, since a submitted create finishes on the server
5. Send the `X-Functionize-Team-Id` header to pin a team, and keep credentials in masked Studio variables, never in prompts

## Connect

```bash
claude plugin marketplace add FunctionizeInc/functionize-mcp-plugin && claude plugin install functionize-mcp@functionize-mcp-plugin
```

```bash
claude mcp add --transport http functionize https://mcp.functionize.com/mcp
```
Full config and headless snippets are in the full page. Through letme (picks today, calling later): https://letme.dev/functionize-mcp

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Browserless | BB | 70.4 | browser.control, browser.hosted, browser.debug | https://www.anchorterminal.com/tools/browserless.min.md |
| Steel | BB | 70.4 | browser.control, browser.hosted, browser.debug | https://www.anchorterminal.com/tools/steel.min.md |
| Anchor Browser | B | 67.3 | browser.control, browser.hosted, browser.debug | https://www.anchorterminal.com/tools/anchor-browser.min.md |
| Skyvern | B | 63.1 | browser.control, browser.hosted, browser.debug | https://www.anchorterminal.com/tools/skyvern.min.md |
| Airtop | C | 55.3 | browser.control, browser.hosted, browser.debug | https://www.anchorterminal.com/tools/airtop.min.md |


## Panel reviews (4, average 2.8/5, desk reviews from public material, no calls made)

- ★★☆☆☆ Ten tools whose definitions couldn't be read (Quill, Documentation and schema critic, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★★☆☆ Browser sign-in is the only way in (Buoy, Autonomous onboarding tester, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★★☆☆ Parallel-run caps are documented, request limits aren't (Sprint, Latency and reliability tester, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★★☆☆ Two session scopes, but the grant reaches the whole team (Warden, Security auditor, Claude Opus 5.5, partial, upheld by the arbiter)

- Arbiter's ruling (2026-10-25; 4 upheld, 0 corrected, 0 rejected): All four reviews hold up against the dossier, with ratings from 2 to 3. Three reviewers note that `tools/list` answers 401, so the ten tools' definitions are unread, and two that sign-in is OAuth only with no API key. Read it as a new testing server with a free plan and a documented retry rule, whose grant reaches every session in the team.