Anchor, which provides browser infrastructure for artificial intelligence agents, unveiled Action Index on October 15th, 2026. The tool functions as a repository of previously completed web workflows, enabling agents to execute proven interaction sequences as deterministic code rather than reasoning through each step independently. When page layouts shift, the underlying model intervenes only at those specific moments.

Testing across 300 tasks drawn from the Online-Mind2Web benchmark revealed striking efficiency gains. A smaller-scale model paired with Action Index achieved a 93.3% task completion rate at just $0.016 per task. By contrast, a state-of-the-art model reached 93% completion at $1.57 per task. The entire 300-task evaluation cost $4.87 when using Action Index, versus $472 without it. The same smaller model achieved only 62.7% completion when operating without the new feature.

A parallel evaluation examining 120 tasks demonstrated additional performance metrics. Action Index decreased token consumption at the 90th percentile by 74%, reduced the number of steps by 75%, and cut wall-clock execution time by 61%. Performance improvements extended to open-source models as well: two prominent open-weight models improved from 67.7% to 88.7% and from 58.7% to 84.7% respectively.

Agents today pay to re-solve problems that are already solved. 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.

Idan Raman, co-founder and CEO of Anchor

Availability and Integration

Action Index launches immediately for all Anchor users, including those on the free tier. The feature integrates through Anchor's MCP server at https://api.anchorbrowser.io/v1/mcp and via the REST API. The system maintains compatibility with any MCP-compatible agent or model provider. Full technical documentation is accessible at https://docs.anchorbrowser.io/advanced/mcp.

About Anchor

Anchor operates as a browser infrastructure platform designed specifically for AI agents. The company supplies agents with secure, dependable cloud-based browsers capable of logging into websites, navigating content, and executing tasks across any web property—including those lacking public APIs.

Source: The Next Web