Arcade.dev launches Tool Recommendation to cut AI agent tool-discovery costs

Arcade.dev has introduced Tool Recommendation, a feature designed to help AI agents find the right actions without repeatedly loading lengthy lists of software functions into their conversations. The company says the approach can cut tokens devoted to tool discovery by up to 91%, with modeled input-cost savings reaching 37%. In a separate test of complete workflows, Arcade reported a 7.6% reduction in combined input/output token bills.

Arcade.dev provides a control layer that helps AI agents operate inside business software while managing permissions, credentials and audit records. The new recommendation system lets an agent describe its immediate need, then searches and ranks Arcade's catalog of more than 8,000 tools before returning a shortlist and executing the selected action. The company says Tool Recommendation works through MCP and can be enabled at its gateway without rebuilding existing integrations.

The feature became available to customers of Arcade Cloud, the company's fully managed SaaS platform, on October 15, 2026, at no separate feature charge. Standard platform and usage fees still apply to paid plans. Arcade has not published pricing or a timeline for a separate enterprise rollout. The company also noted that the savings figures are based on modeled scenarios or limited workflow tests, not independently verified customer billing results.