메뉴
Use and maintain the fetch-mcp networking server. Applies when fetching URLs, optimizing JSON, using smart_fetch, browser_fetch, web_search, css_query, CLI commands, MCP tools, Playwright rendering, or handling HTTP-client-blocked pages and CAPTCHA.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/micaelmalta/fetch-mcp --skill fetch-mcp명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SKILL.md 표시 중
SOC 직업 분류 기준
| name | fetch-mcp |
| description | Use and maintain the fetch-mcp networking server. Applies when fetching URLs, optimizing JSON, using smart_fetch, browser_fetch, web_search, css_query, CLI commands, MCP tools, Playwright rendering, or handling HTTP-client-blocked pages and CAPTCHA. |
Use this skill when the user wants to fetch web pages, search the web, reduce JSON/token size, debug fetch-mcp behavior, or choose between smart_fetch, browser_fetch, css_query, web_search, and optimize_json.
Use smart_fetch first for normal URLs:
uv run fetch-mcp smart_fetch https://example.com
uv run fetch-mcp smart_fetch https://api.github.com/orgs/python/repos --jsonpath '$[*].name'
Use browser_fetch when a page needs JavaScript rendering or blocks simple HTTP clients:
uv run fetch-mcp browser_fetch https://example.com
uv run fetch-mcp browser_fetch https://example.com --selector main
Use headed browser mode only for human-in-the-loop login or CAPTCHA:
uv run fetch-mcp browser_fetch https://example.com --headed --wait-ms 30000
Use optimize for JSON already available on stdin:
curl -s https://api.github.com/orgs/python/repos | uv run fetch-mcp optimize
cat response.json | uv run fetch-mcp optimize --jsonpath '$[*].name'
Use report to inspect cumulative savings:
uv run fetch-mcp report
Do not bypass CAPTCHA, login walls, paywalls, robots restrictions, or anti-bot systems with stealth headers, proxy rotation, CAPTCHA solvers, or fingerprint evasion.
If smart_fetch returns 403 or anti-bot content, switch to browser_fetch only to render the page like a normal browser. If a CAPTCHA or login appears, use --headed and require the human to complete it manually before extraction.
The server is split across fetch_mcp/ modules:
fetch_mcp/server.py — MCP tool registrations (@mcp.tool)fetch_mcp/cli.py — CLI dispatch and main()fetch_mcp/json_optimizer.py — pruning pipeline and schema-first modefetch_mcp/html.py — HTML→Markdown conversionfetch_mcp/http.py — HTTP client, SSL, Chrome detectionfetch_mcp/savings.py — savings loggerfetch_mcp/_resolve.py — optimize_json input resolutionKeep smart_fetch as the fast HTTP path using httpx.
Keep browser_fetch as the slower Playwright/Chrome path. It should be optional at runtime when possible, and its CLI help should make headed mode discoverable.
For HTML conversion, use the local _html_to_markdown() helper from fetch_mcp/html.py so output stays consistent across smart_fetch, css_query, and browser_fetch.
For large JSON, preserve schema-first behavior by default and use JSONPath for drill-down.
After changes, run:
uv run python -m pytest tests/ -v
uv run ruff check .
uv run fetch-mcp -h
uv run fetch-mcp smart_fetch https://example.com
uv run fetch-mcp browser_fetch https://example.com --max-chars 2000
printf '{"items":[{"name":"a"},{"name":"b"}]}' | uv run fetch-mcp optimize