RobotMCP
The agent finds Robot Framework keywords, runs steps live on a site, API or app and saves a ready .robot suite.
- 118 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The agent runs Robot Framework steps in your environment: browser, mobile device, database and desktop.
- Keywords can change data in the database and on tested sites.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/robotmcpDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs Robot Framework steps in your environment: browser, mobile device, database and desktop.
- Keywords can change data in the database and on tested sites.
This is third-party code. Review the repository files before installing.
What it does
RobotMCP (`rf-mcp`) gives an agent access to Robot Framework. The agent finds suitable keywords, runs steps live through Browser, Selenium, Appium, Requests, a database or the desktop, sees the real result and only after a passing run assembles a clean `.robot` suite. From a plain-language description the server builds BDD-style suites and data-driven templates. `robotmcp install` writes the configuration into Claude Code, Codex, GitHub Copilot, opencode, Gemini CLI, Kilo Code, goose and Cursor, and `robotmcp doctor` shows which environment the server will see. There is a Django dashboard, instruction templates and optional persistent semantic memory.
Who it is for. QA engineers and Robot Framework developers who want an agent to write verified scenarios instead of invented ones.
Good fit when
- A login and checkout scenario is needed, and the agent picks locators on the live page.
- API tests on RequestsLibrary are written from a task description.
- The project already has its own keywords, and `robotmcp install` from the project directory attaches them to the server.
- Tests are wanted in Given/When/Then style or as a data-driven template.
Not a fit when
- Tests are not written in Robot Framework.
- The project pins Robot Framework 6.x: the server routes the run to the attach bridge and does not execute tests on another version.
- A single one-off request is enough, with no suite to build.
Example request
Create a test suite for https://www.saucedemo.com/: log in, add two items to the cart, check out and verify the success message. Use Selenium Library and run steps one at a time.MCP
- Transport
- stdio, http
- Authentication
- not required
Environment variables
ROBOTMCP_MEMORY_ENABLED- The value true enables persistent semantic memory.
Limitations
Python 3.10 or newer is required. The server imports Robot Framework libraries into its own process and runs tests there, so for a project with its own libraries run `robotmcp install` from its directory. Browser Library needs Playwright browsers (`robotmcp init --browsers`) and Node.js. Desktop (PlatynUI) is not part of `[all]`, installs separately with `--prerelease=allow` on Python 3.12 or newer and runs on Linux with glibc 2.34 or newer, macOS on Apple Silicon and Windows.
How to disable. Run `robotmcp uninstall`, a safe and reversible removal of the server entries from agent configurations. Remove the tool itself with `uv tool uninstall rf-mcp`.
FAQ
How do I install RobotMCP in Claude Code?
Run `uv tool install "rf-mcp[all]"` and then `robotmcp install`, which writes the configuration into detected agents. By hand: `claude mcp add rf-mcp -- uvx rf-mcp`.
Which extras should I pick?
`api` installs RequestsLibrary with no extra steps, `web` adds SeleniumLibrary and Browser, `mobile` adds AppiumLibrary and `database` adds DatabaseLibrary. `all` includes everything except `desktop`.
Why does the server not see my libraries?
It runs in its own environment. Run `robotmcp install` from the project directory: it finds the project environment and attaches its libraries. Verify with `robotmcp doctor -C .`.
Which agents does `robotmcp install` support?
Claude Code, OpenAI Codex, GitHub Copilot, opencode, Gemini CLI, Kilo Code, goose and Cursor. Each gets configuration in its own format and other servers are left untouched.
Is there a Docker image?
Yes: `ghcr.io/manykarim/rf-mcp` for headless runs and `ghcr.io/manykarim/rf-mcp-vnc` with a VNC desktop. Port 8000 serves MCP over HTTP and 8001 opens the dashboard.
Discussion
No questions or reviews yet.
Related
MCP serversWebdriverIO MCP: browser and mobile app automation
webdriverio-mcp
The official MCP from the WebdriverIO team drives Chrome, Firefox, Edge, Safari, Electron, iOS and Android through one tool set
MCP Interviewer checks an MCP server before an agent connects: provider limits, tool tests and a Markdown report.
kogiQA MCP gives an agent a browser that finds page elements without selectors, for debugging web apps and writing e2e tests
An MCP server from Mozilla: an agent drives Firefox through WebDriver BiDi, reads the console and network, takes page snapshots and profiles loading
BrowserStack MCP Server: testing on real devices
BrowserStack MCP Server
The official BrowserStack MCP server: real devices, automated tests and test cases driven by an agent from your editor
A community Playwright MCP server: browser automation, screenshots, device emulation and test code generation
README in short
The server plans and runs steps first and builds the suite after success. Responses can be trimmed for small models through tool profiles, state deltas and artifact externalization.