Skip to content

Releases: us/crw

v0.21.0

Choose a tag to compare

@us us released this 06 Jul 20:43

0.21.0 (2026-07-06)

⚠ BREAKING CHANGES

  • api: error responses now use errorCode instead of error_code. The managed API already strips the field, so managed clients are unaffected; self-hosted clients parsing error_code should read errorCode.

Features

  • cli: add crw bench FRAMES harness (fb78bcb)
  • cli: bench A/B flags, concurrency, crash-safe writes (7eb059e)
  • core: add evidence & provenance primitives (4a638a0)
  • core: expose sourceHash on scrape responses (b0d89f9)
  • extract: prompt-based extraction and full meta-tag metadata (6453204)
  • proxy: retry with default country on residential CONNECT tunnel failure (8aafdc8)
  • search: thread per-request country to result-page scraping (be3b0d1)
  • server: add /firecrawl/* compat namespace, own /v1 as native API (d9cb3b8)

Bug Fixes

  • api: rename error_code response field to errorCode (e4df682)
  • cli: repair clap arg conflict that panicked debug builds (f2a6b77)
  • docs: add redirect stubs for pre-flatten legacy doc URLs (a1f29dc)
  • extract: unify untrusted-content fencing; nonce-fence the change judge (192b9d7)
  • renderer: cap full-page screenshot height to avoid OOM (c5f555d), closes #161
  • renderer: don't browser-render thin pages that ship no JS (f42f14a)
  • scrape: report anti-bot/challenge pages as blocked, not success (d06f051)
  • search: charset-aware scrape decode + per-result error + neutral answer warnings (a9870e2)

v0.20.0

Choose a tag to compare

@us us released this 01 Jul 11:43

0.20.0 (2026-07-01)

Features

  • map: escalate anti-bot-gated sitemaps through the JS renderer (2634a7e)
  • map: fix URL discovery on hard sites + configurable discovery limit (f5fe1f4)

Bug Fixes

  • renderer: flat 1-credit cost for every renderer (35a604c)

v0.19.0

Choose a tag to compare

@us us released this 28 Jun 15:04

0.19.0 (2026-06-28)

Features

  • renderer: proxy-retry on origin rate-limit (429) (da7ecda)
  • renderer: relaxed-TLS fallback for cert-broken origins (13d39f6)

Performance

  • engine: offload HTML extraction off the async reactor (1683153)

v0.18.3

Choose a tag to compare

@us us released this 23 Jun 12:00

0.18.3 (2026-06-23)

Bug Fixes

  • renderer: strip "Mozilla" UA prefix for lightpanda (it rejects it) (fda6139)

v0.18.2

Choose a tag to compare

@us us released this 23 Jun 10:51

0.18.2 (2026-06-23)

Bug Fixes

  • renderer: send a modern UA on the CDP path (setUserAgentOverride) (ffcf564)

v0.18.1

Choose a tag to compare

@us us released this 23 Jun 05:04

0.18.1 (2026-06-23)

Bug Fixes

  • npm: ship bin/install.js + bin/agents.js in the crw-mcp package (437dbc7)

v0.18.0

Choose a tag to compare

@us us released this 23 Jun 01:45

0.18.0 (2026-06-23)

Features

  • extract: optional reasoning_effort config field (e677190)
  • install: auto-launch crw setup after install + conversion-tuned copy (5cc35f9)
  • mcp: add install command that wires skill + MCP for all agents (4b41621)

Bug Fixes

  • ci: rebase-retry docs-sync push + sync 0.17.0 changelog (597aa90)
  • install: default to the crw CLI + drop the musl/glibc gate (9958273)
  • install: drop the GB figure from local, surface 500 credits everywhere (f997aa0)
  • mcp: clarify crawl/parse jsonSchema arg for MCP clients (91be697)
  • mcp: static musl linux binaries + correct claude mcp add docs (75333d6)
  • server: make crw- model-prefix boot guard opt-in (5cb867c)
  • stealth: bump UA + Sec-Ch-Ua from Chrome 131 to 150 (695863c)

v0.17.0

Choose a tag to compare

@us us released this 21 Jun 21:52

0.17.0 (2026-06-21)

Features

  • renderer: add opt-in Camoufox stealth renderer tier (744beda)
  • renderer: conditional hedge + event-driven readiness for p90 (9065007)
  • scrape: add screenshot output format via CDP capture (61e03e7), closes #161
  • sdk: add Research API methods to TS + Python SDKs (3a2f710)
  • search: add Firecrawl-compatible research API engine layer (ba1a87c)
  • search: overlap query-expansion scrape with original (C1) (4f6147e)
  • skills: add crw agent skill set (0ddbe01)
  • skills: publish crw-research agent skill + docs install command (4638715)

Bug Fixes

  • docs: render :::tabs and :::callouts in prerendered pages (f3a495a)
  • map: render SPA shells during URL discovery (0ec4bf9), closes #166
  • mcp,sdk: drop phantom search country param, export CrwApiError (58b8e5c)
  • pdf: bound sandbox child address space to prevent false pdf_too_large (06acb83)
  • proxy: normalize empty CRW_CRAWLER__PROXY to None (#154) (b3d0fe9)
  • scrape: capture screenshot outside the nav-budget race (4021b50)
  • search: resolve arXiv inspect via Semantic Scholar (70126c6)

Performance

  • search: research concurrency 4->8, cache cap 20k->3k (e3a6ac3)

v0.16.0

Choose a tag to compare

@us us released this 14 Jun 11:01

0.16.0 (2026-06-14)

Features

  • mcp: optimize MCP server for context, weight, and conformance (aac7999)
  • proxy: add proxy list + rotation primitives and HTTP-path rotation (776e9fb)
  • proxy: rotate the JS/Chrome (CDP) path per request (422ac09)
  • proxy: v2 BYOP plumbing + honest docs + verification harness (0983ba3)

Bug Fixes

  • extract: stop doubling /v1 in structured-extraction chat URL (d8b8ebc)
  • extract: unify Anthropic structured URL, stop /v1/messages doubling (90ce3dd)
  • proxy: accept snake_case proxy_list alias on v1 ScrapeRequest/CrawlRequest (a8e7b71)
  • proxy: CLI crawl/map --proxy reaches the JS/CDP tier (round-4 review) (6ee7175)
  • proxy: resolve review findings (IP-leak/correctness hardening) (72e0486)
  • proxy: route /map discovery through the rotator (round-2 review) (5ee06cf)
  • proxy: route crawl robots/sitemap egress through the rotator (4835e7d)

v0.15.2

Choose a tag to compare

@us us released this 12 Jun 12:08

0.15.2 (2026-06-12)

Bug Fixes

  • mcp: npm launcher downloads binary when platform pkg missing (4ae1aa6)