Mcp Fetch
Fetch public URLs and APIs into clean markdown, text, HTML, or JSON for the agent with SSRF guards so research and integrations stay safer.
Overview
mcp-fetch is an Idea-phase MCP server that SSRF-guards URL and REST fetches and returns markdown, text, HTML, or JSON for AI agents.
What is this MCP server?
- SSRF-safe URL fetching for agent-driven web research
- Normalized outputs: markdown, plain text, HTML, and JSON
- Guarded REST-style calls for lightweight API probes
- Reduces accidental internal-network requests from agents
- Stdio MCP @infoinlet/mcp-fetch v0.1.1 on npm
- Server version 0.1.1
- npm package @infoinlet/mcp-fetch
- stdio transport
What problem does it solve?
Unrestricted agent fetch tools can hit internal URLs or return noisy HTML that breaks research and invites security mistakes.
Who is it for?
Builders doing early market and API discovery who want fetch in MCP without opening full browser automation.
Skip if: Logged-in scraping, complex JavaScript-rendered apps, or penetration testing—use dedicated browser or security tooling instead.
What do I get? / Deliverables
Your agent gets cleaner, safer web and API snapshots for competitor and documentation research inside the IDE.
- Agent-ingested page content in markdown or text form
- Structured JSON snapshots from allowed API URLs
- Safer fetch behavior versus unconstrained HTTP tools
Recommended MCP Servers
Journey fit
Idea-phase research depends on pulling competitor pages, docs, and API responses before you commit to a build plan. Research is the canonical shelf for guarded HTTP fetch that replaces risky unrestricted wget-style agent tools.
How it compares
SSRF-aware fetch MCP, not Playwright-style browser automation or a raw curl skill without guardrails.
Common Questions / FAQ
Who is mcp-fetch for?
Solo builders and agents users who need safe HTTP fetch and light REST probes during research and integration experiments.
When should I use mcp-fetch?
Use it in idea and research when pulling public docs, competitor pages, or JSON APIs before you harden production crawlers.
How do I add mcp-fetch to my agent?
Install @infoinlet/mcp-fetch from npm, register it as a stdio MCP server in your client, and follow any URL policy notes in the marketplace repo.