Anchor Blog · Oct 17, 2026 · 10:17 PM

Anchor's Action Index lets web agents replay proven workflows as code, cutting per-task cost ~98x

Action Index stores web workflows that already succeeded and replays them as deterministic code, calling the model only on steps where the page changed; in Anchor's Online-Mind2Web test, gpt-5.6-luna with it hit 93.3% at $0.016 per task versus claude-opus-5's 93.0% at $1.57. It's on every Anchor plan including free, via search_action_index/execute_action_index tools on https://api.anchorbrowser.io/v1/mcp or the REST API.

Read at anchorbrowser.io

More news

View all

Perplexity Computer adds REA's /reverse-engineer skill, free up to 10,000 credits through Oct 31

Computer can now use morluto's open source REA toolkit to inspect how an app works without its source code and build a working version; tasks started with /reverse engineer are free through October 31, up to 10,000 credits per user. Perplexity showed five demo rebuilds, including a Figma style editor with multiplayer editing and a LibreOffice Calc style spreadsheet…

Perplexity (X) 1 hour ago

Kondo's WhatsApp MCP server lets Claude, ChatGPT, and Cursor read and draft chats but never send

Announced Oct 17, adding https://mcp.trykondo.com/mcp lets an MCP client search WhatsApp chats, apply labels, snooze follow ups, archive, and save reply drafts, read from the Kondo desktop app on your own computer while it's open. There's no send tool at all, and MCP is on Kondo's Business and Enterprise plans only…

Kondo 3 hours ago

Replit makes OpenAI the default for Replit Agent, says it no longer works with Anthropic

At TechCrunch Disrupt, CEO Amjad Masad said Replit stopped working with Anthropic over cost, partnership, and performance, later clarifying that users can still pick Claude but OpenAI's models are now Replit Agent's default; people familiar with the deal told TechCrunch the partnership is still in place. Masad also said enterprise contracts grew from about 2% of revenue last year to about 15% this year…

TechCrunch 4 hours ago

DeepSeek opens standalone Search and Fetch APIs for agents at $0.75 per 1,000 calls

As of Oct 17, any DeepSeek API key can call POST /v1/tools/search (ranked web results with quality and authority scores) and POST /v1/tools/fetch (one page as Markdown), each billed only on successful calls and capped at 5 requests per second per account. Server side search and fetch from DeepSeek Harness, Claude Code, and Codex on the DeepSeek API stays free until Oct 31, 2026, 16:00 UTC…

DeepSeek API Docs 5 hours ago

DeepSeek-V4.1-Flash-1017 ships MIT open weights with big Terminal-Bench and DeepSWE gains

DeepSeek's Oct 17 update to V4.1 Flash (763B params, 1M context, MIT license) raises Terminal Bench 4.0 from 31.2 to 42.3 and DeepSWE v1.1 from 74.2 to 77.0 in its own tests, while finishing tasks faster with shorter outputs. It's on DeepSeek's API and app, plus Vercel AI Gateway as deepseek/deepseek v4.1 flash 1017…

Hugging Face (deepseek-ai) 6 hours ago

Spotted something we missed? Start a thread.