How do I use AgentFetch with Continue.dev?

Use AgentFetch with Continue.dev by adding an MCP server entry to ~/.continue/config.json under the experimental.modelContextProtocolServers array: {"name": "agentfetch", "transport": {"type": "stdio", "command": "npx", "args": ["-y", "@agentfetch/mcp"]}}. Restart your IDE (Continue runs as a VS Code or JetBrains extension), reload the Continue panel, and fetch_url plus extract_json appear in the agent's tool list. Continue added MCP support in late 2024 and it remained behind the experimental flag through 2025; check the latest Continue docs in case the config path moved out of experimental in your version. Continue is popular with self-hosted-model users (Ollama, llama.cpp, vLLM), and AgentFetch's markdown output is even more valuable here because local models typically have smaller context windows (8k-32k) than Sonnet/Opus — a 50KB raw HTML page won't even fit, while a 5KB markdown version leaves room for the model to reason. The same AGENTFETCH_API_KEY env var unlocks paid tier limits. Verify by running /tools in the Continue chat — fetch_url should be listed.