What are AgentFetch's rate limits?

AgentFetch's rate limits are tiered: free tier allows 1,000 fetches/month and 10 requests/minute, sufficient for hobby projects, demos, and individual Claude Desktop or Cursor use. Paid tiers (Starter, Pro, Team) raise both limits — Starter at 10k fetches/month and 60 req/min, Pro at 100k/month and 300 req/min, Team at custom volumes with dedicated proxy pools. Internally, AgentFetch also applies per-host rate budgets to protect upstream publishers — typically 1 request per second per host with burst tolerance, configurable on Pro+ tiers. When you hit a limit, the tool returns a structured error (rate_limited, with retry_after_seconds) that the agent can reason about and Claude/Cursor surface as a tool error — better than silent failure. For burst workloads (e.g., ingesting 500 URLs for a one-off RAG load), AgentFetch supports priority=batch which queues requests, runs them at the configured QPS, and streams results — total time scales linearly but stays inside rate budgets. For self-hosted deployments, you set your own limits via env vars (AGENTFETCH_GLOBAL_QPS, AGENTFETCH_PER_HOST_QPS). The hosted free tier resets monthly on the first of the calendar month; paid tiers reset on the billing anniversary. Check current quota at any time via the account_info MCP tool or the dashboard at https://www.agentfetch.dev/dashboard.