Skip to content

occam_crawl

Status: EXPERIMENTAL · enable with OCCAM_CRAWL_MCP=1
Contract: runtime tools/list when the flag is set · ADR-0018

Same-domain crawl: discover links from a seed (sitemap then homepage fallback), then transcode up to max_pages. Pass resume_token from a prior response to continue the queue. Optional jsonl of page rows.

Not a CAPTCHA unlocker. Soft challenges still follow probe → transcode → optional session_profile.

Parameters

Name Required Default Notes
url if no resume — Seed URL
resume_token if no url — Caller-held queue from prior crawl
max_pages no 8 1–32
max_links no 32 Discovery budget (1–64)
map_source no sitemap sitemap or homepage
backend_policy no http_then_browser
focus_query / fit_markdown / per_url_max_tokens no prune defaults Per-page compile
session_profile / playbook_policy no — / auto
include_jsonl no true