Agentfetch
Fetch web pages and PDFs into your agent context with automatic routing and token budgets so research does not blow the context window.
Overview
Agentfetch is an MCP server for the Idea phase that performs token-budgeted web and PDF fetches with automatic routing across Jina, FireCrawl, Trafilatura, and PDF backends.
What is this MCP server?
- Token-budgeted web fetch designed for AI agent sessions
- Auto-routes across Jina Reader, FireCrawl, Trafilatura, and PDF handling
- Optional JINA_API_KEY and FIRECRAWL_API_KEY with documented free tiers
- Optional REDIS_URL for response caching
- PyPI agentfetch-mcp v1.0.1 with stdio transport
- Version 1.0.1; PyPI identifier agentfetch-mcp; stdio transport
- Documented optional tiers: Jina ~1M tokens/mo; FireCrawl 500 free credits
- Env vars: JINA_API_KEY, FIRECRAWL_API_KEY, REDIS_URL
What problem does it solve?
Manual copy-paste and ad-hoc scraping eat context and break on JS or PDF pages when you are trying to research a market in one agent session.
Who is it for?
Solo builders who research competitors and docs daily inside Claude Code and want one MCP fetch layer with optional Redis cache.
Skip if: Teams that only need gov contract NAICS feeds or Stripe agent checkout with no web reading.
What do I get? / Deliverables
After you add agentfetch-mcp and optional API keys, your agent gets routed, budget-aware page text without you wiring each fetch provider yourself.
- Clean extracted text from URLs and PDFs within agent token budgets
- Automatic provider routing without per-site fetch scripts
Recommended MCP Servers
Journey fit
How it compares
Budgeted multi-provider fetch MCP, not a single-site browser skill or penetration-testing crawler.
Common Questions / FAQ
Who is Agentfetch for?
Indie builders and agent users who need reliable, token-conscious web and PDF text during research and reference lookups from the IDE.
When should I use Agentfetch?
Use it during idea research—or anytime you need live URLs in context—when JS rendering, PDFs, or provider choice would otherwise block quick validation.
How do I add Agentfetch to my agent?
Install agentfetch-mcp from PyPI, configure stdio in your MCP client, and set JINA_API_KEY and optional FIRECRAWL_API_KEY and REDIS_URL.