mirror of
https://github.com/getmaxun/maxun.git
synced 2026-10-27 19:27:48 +08:00
The MCP worker could only list and run existing robots. It now exposes Maxun's core capabilities as one-shot tools, plus run management: - maxun_scrape: scrape one page (markdown, html, text, links, summary, screenshots) with optional smart queries - maxun_crawl: crawl a site with scope, depth, include/exclude paths - maxun_search: web search in discover or scrape mode, with time range - maxun_agent: build an AI-mode extraction robot from a prompt and run it - maxun_cancel_run, maxun_get_run_diff: cancel runs, inspect changes Existing tools are renamed with the maxun_ prefix (get_run_details is now maxun_get_run). One-shot tools reuse a robot when called again with the same arguments, so repeat calls build a run history on one robot. Page content is returned as plain text blocks instead of escaped JSON, API errors now include the server's message, and LLM settings for agent, summary and smart queries are read from MAXUN_LLM_* env variables. The SDK execute endpoint now records runs started with x-run-source: mcp as MCP runs and accepts wait: false to return as soon as the run is queued. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Rx3RjPpi36oz75AQnmUy2N Signed-off-by: Karishma Shukla <karishmashuklaa@gmail.com>