ctfd-mcp
ctfd-mcp
Documentation
CTFd MCP server (user scope)
MCP server that lets a regular CTFd user list challenges, read details, start/stop dynamic docker instances, and submit flags.
Requirements
- Python 3.13 (managed by `uv`).
- Environment variables (choose one auth method):
- `CTFD_URL` (e.g. )
- `CTFD_TOKEN` (user token, not admin) or `CTFD_SESSION` (session cookie if tokens are disabled).
- `CTFD_CSRF_TOKEN` (optional, only if the server/plugin requires CSRF for ctfd-owl).
You can store them in a `.env` file in the repo root:
CTFD_URL=https://ctfd.example.com/
CTFD_USERNAME=your_username
CTFD_PASSWORD=your_password
# or, if you prefer to use a token:
# CTFD_TOKEN=your_ctfd_api_token_here
# or, if tokens are disabled:
# CTFD_SESSION=your_session_token_here
# and, if the owl plugin enforces CSRF:
# CTFD_CSRF_TOKEN=your_csrf_token_hereInstall
- From PyPI (recommended): `uvx ctfd-mcp --help`
- From source checkout (no install): `uvx --from . ctfd-mcp --help`
Run MCP server (stdio)
# installed from PyPI
uvx ctfd-mcp
# from local checkout
uvx --from . ctfd-mcpCursor and Claude MCP config example
{
"mcpServers": {
"ctfd-mcp": {
"command": "uvx",
"args": ["ctfd-mcp"],
"env": {
"CTFD_URL": "https://ctfd.example.com",
"CTFD_TOKEN": "your_user_token"
}
}
}
}Codex MCP config example
[mcp_servers.ctfd-mcp]
command = "uvx"
args = ["ctfd-mcp"]
[mcp_servers.ctfd-mcp.env]
CTFD_URL = "https://ctfd.example.com"
CTFD_TOKEN = "your_user_token"Exposed tools
- `list_challenges(category?, only_unsolved?)` — list visible challenges, optional category/unsolved filter.
- `challenge_details(challenge_id)` — description (HTML + `description_text`), metadata, attachment URLs, solved status.
- `submit_flag(challenge_id, flag)` — attempt a flag; returns status/message.
- `start_container(challenge_id)` — unified start; auto-detects dynamic_docker, ctfd-owl or k8s `/api/v1/k8s`.
- `stop_container(container_id?, challenge_id?)` — unified stop; whale can be stopped with just `container_id`, owl/k8s need `challenge_id`.
Attachments are returned as absolute URLs in `files`; the client/host can fetch them directly.
MCP resources
- `resource://ctfd/challenges/{challenge_id}` — markdown snapshot of a challenge (metadata, description, attachment URLs, connection info if present).
Error handling
- Missing env/config -> clear MCP error.
- 401/403 -> auth failed, check token or session cookie.
- 404 -> not found (or dynamic container API missing).
- 429 -> rate limited (Retry-After if present).
- Other HTTP/API errors -> surfaced as MCP errors with CTFd message/status.
Notes and troubleshooting
- Dynamic containers require the ctfd-whale (dynamic_docker) plugin on the target CTFd; otherwise `/api/v1/containers` returns 404.
- Owl challenges (`dynamic_check_docker`) use a different endpoint: `/plugins/ctfd-owl/container?challenge_id=`. They usually require a session cookie, and some setups require a CSRF token; set `CTFD_CSRF_TOKEN` if needed.
- Some events expose Kubernetes-backed instances at `/api/v1/k8s/{get,create,delete}` with multipart form data; the client will try these when the challenge type includes `k8s` (or when a dynamic_docker endpoint is missing).
- If the server redirects you to `/login` (302) when using a token, switch to a browser session cookie: set `CTFD_SESSION` from the `session` cookie after logging in.
- The client now supports logging in with `CTFD_USERNAME` and `CTFD_PASSWORD`; these fields take precedence over stale tokens/sessions.
- Auth priority: username/password first, then token, then session cookie. Lower-priority credentials are ignored when a higher-priority option is present.
Support / feedback
If something breaks or you have questions, reach out:
- Telegram: @ismailgaleev
- Jabber: ismailgaleev@chat.merlok.ru
- Email: umbra2728@gmail.com
Testing
- Run `uv run pytest`.
- Timeouts are configurable via env: `CTFD_TIMEOUT` (total), `CTFD_CONNECT_TIMEOUT`, `CTFD_READ_TIMEOUT` (seconds). Defaults are 20s total / 10s connect / 15s read.
Development
- Dev dependencies: `uv sync --group dev`
- Lint/format: `uv run ruff check .` and `uv run ruff format .`
- Tests: `uv run pytest`
- Pre-commit: `uv run pre-commit install` (see `CONTRIBUTING.md`)
License
Apache-2.0. See `LICENSE`.
Frequently asked questions
What is ctfd-mcp?
ctfd-mcp is ctfd-mcp
How do I install ctfd-mcp?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is ctfd-mcp open source?
Yes — it is hosted on GitHub at https://github.com/umbra2728/ctfd-mcp and has 25 stars.
Related MCP tools
Cognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory across sessions with a self-hosted knowledge graph engine.
Automate browser based workflows with AI
Hindsight: Agent Memory That Learns
A privacy-first app that strips AI watermarks from content you own.
Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.
The power of Claude Code / GeminiCLI / CodexCLI + [Gemini / OpenAI / OpenRouter / Azure / Grok / Ollama / Custom Model / All Of The Above] working as one.
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP