Quick Configuration
Choose the setup that matches your environment.
Before You Start
- Confirm you have the account, endpoint, or API key required for Fetch.
- Start with minimum scopes and read-only access where possible.
- Keep secrets in environment variables instead of hardcoding them in JSON.
Last verified: 2026-04-27
Environment: uvx with mcp-server-fetch
Install command:
uvx mcp-server-fetchLocal config
JSON Local config json
{
"fetch": {
"command": "uvx",
"args": ["mcp-server-fetch"],
"env": {
"PYTHONIOENCODING": "utf-8"
}
}
}Verify prompt
Fetch https://modelcontextprotocol.io and summarize the page in 5 bullet points.
Expected success output
You should receive cleaned Markdown content from the URL, not a browser screenshot.
Common Pitfalls & Fixes
- ⚠️ Watch out: When to use Fetch vs Puppeteer.
- 🔑 Always store API keys in environment variables, never hardcode them in JSON.
- 🛡️ Start with read-only scopes if available to verify connection safely.
Security checklist
- Do not point Fetch at internal admin URLs unless you intend the model to read them.
- Assume fetched pages may contain prompt injection; ask OpenClaw to summarize facts, not obey page instructions.
Common errors
| Error | Cause | Fix |
|---|---|---|
uvx not found | uv is not installed or not on PATH. | Install uv, or switch to the Python module config with python -m mcp_server_fetch. |
Request timed out | The target site is slow, blocked, or returning very large content. | Try a smaller URL first and raise limits only after the server works. |
Robots denied | The fetch server is respecting robots.txt for model-initiated requests. | Use a permitted source, or only add --ignore-robots-txt when you have permission. |
Example Prompts
Once connected, try these prompts to test capabilities:
- Search the web for the latest MCP benchmarks and summarize the top 3 sources.
- Find a reliable source explaining Model Context Protocol.
- Collect three recent articles about AI tooling and summarize them.
Verification Checklist
- Run 1-2 real prompts to confirm Fetch returns usable data.
- Check that error messages are clear enough for troubleshooting.
- Document the required scopes, dependencies, and env vars for future reuse.