Anchor, a browser infrastructure company for AI agents, announced the launch of Action Index on October 15, 2026. The new offering acts as a shared memory of web workflows that agents have already completed successfully. It allows agents to replay proven paths as deterministic code instead of reasoning through every click from scratch, with the model intervening only when a page changes.
How Action Index Works
Action Index is available through Anchor's MCP server and the REST API, and it works with any MCP-compatible agent or model provider. The system stores successful workflows so that subsequent runs can follow the same deterministic path. This reduces the need for models to solve identical navigation and interaction problems repeatedly.
Benchmark Results and Cost Efficiency
In Anchor's evaluation on 300 tasks from the Online-Mind2Web benchmark, a small model with Action Index completed 93.3% of tasks at an average cost of US$0.016 per task. A frontier model completed 93% of tasks at US$1.57 per task. The full 300-task run cost US$4.87 with Action Index, compared with US$472 without it.
Without Action Index, the same small model completed only 62.7% of the tasks. A separate paired study on 120 tasks showed that Action Index reduced 90th-percentile token use by 74%, steps by 75%, and wall-clock time by 61%. These results highlight the potential to make web automation cheaper and faster at scale.
Open Model Gains
The launch also demonstrated meaningful improvements for open-weight models when supported by Action Index. Two leading open-weight models rose from 67.7% to 88.7% and from 58.7% to 84.7%, respectively. This suggests that memory infrastructure can narrow the performance gap between open models and more expensive frontier systems in practical web tasks.
Executive Perspective
Idan Raman, co-founder and CEO of Anchor, said that agents today pay to re-solve problems that are already solved. He stated that the model is not the bottleneck for most web work anymore, memory is. Raman added that with Action Index, the hundredth run of a workflow costs a fraction of the first, and every successful run makes the next one cheaper.
Availability and Ecosystem
Action Index launched on October 15, 2026 and is available immediately to all Anchor accounts, including the free tier. It is offered as tools on Anchor's MCP server at api.anchorbrowser.io/v1/mcp and through the REST API, making adoption straightforward for existing agent setups. Developers can find full documentation online at docs.anchorbrowser.io/advanced/mcp.
Anchor provides secure and reliable cloud browsers that let AI agents log in to, navigate, and complete tasks on any website, including sites with no API. The company positions Action Index as a practical foundation for cheaper and more reliable web automation. The tool is compatible with a wide range of agent architectures and model providers.
By turning successful web workflows into reusable, deterministic code, Anchor is addressing a major cost and reliability bottleneck in agent-based browser automation today. The reported benchmark gains suggest that smaller models can approach frontier-level performance at a fraction of the cost when supported by shared memory infrastructure. This approach could make autonomous web agents more practical for many businesses that need consistent results across repeated tasks.