In rigorous testing across 300 Online-Mind2Web challenges, combining a lightweight artificial intelligence system with a new execution method yielded a 93.3% success rate while spending roughly $0.016 per assignment. Conversely, a state-of-the-art frontier model attained a nearly identical 93% completion rate but required $1.57 per job. Consequently, executing the full battery of 300 operations tallied just $4.87 using the fresh approach, compared to a staggering $472 via traditional techniques. Unassisted by this system, that same modest model managed to resolve merely 62.7% of the evaluated routines.
“Agents today pay to re-solve problems that are already solved,” said Idan Raman, co-founder and CEO of Anchor. “The model isn’t the bottleneck for most web work anymore. Memory is. Action Index means the hundredth run of a workflow costs a fraction of the first, and every successful run makes the next one cheaper.” Industry observers point out that requiring autonomous programs to continually re-evaluate routine online steps places an unnecessary financial and computational strain on software deployments. By shifting the focus toward shared recall, software agents can bypass tedious processing overhead on recurring web interactions.
To tackle these persistent inefficiency challenges, browser infrastructure developer Anchor announced the release of Action Index on October 15th, 2026, in New York, United States. The platform functions as a collective historical repository detailing previously mastered digital routines. Rather than forcing an autonomous system to evaluate every mouse movement and click from scratch, Anchor allows programs to execute established sequences as precise code, invoking generative artificial intelligence solely when site layouts alter. This hybrid execution style ensures that automated web navigation remains both adaptable and highly economical for enterprise application.
A controlled comparative experiment involving 120 specific assignments demonstrated that utilizing this infrastructure diminished 90th-percentile token usage by 74%, decreased necessary operational steps by 75%, and slashed total wall-clock time by 61%. Furthermore, the framework delivered substantial performance improvements to open-source software, elevating one leading open-weight model from 67.7% to 88.7% accuracy and another from 58.7% to 84.7%. Anchor specializes in delivering secure, dependable cloud-based browser environments that empower autonomous systems to authenticate, maneuver, and execute complex duties across internet sites lacking traditional APIs.
Access to the Action Index technology has opened immediately for all registered accounts, encompassing those on no-cost subscription tiers. Developers can integrate the capability via REST API or through tools hosted on Anchor’s Model Context Protocol server at https://api.anchorbrowser.io/v1/mcp, which maintains full interoperability alongside any MCP-supported agent or model provider. Complete technical specifications and implementation guides are hosted at https://docs.anchorbrowser.io/advanced/mcp. Inquiries regarding this release can be sent directly to chief executive officer Idan Raman via idan@anchorbrowser.io.
Leave a Reply