AgentWorksThe operating loop
Build AI workflows that improve with every run.
AgentWorks turns recurring agent work into an observable operating loop: define the outcome, route each step to the right worker, preserve evidence, measure health with Pulse, and improve the next run with human judgment.
Run with the right worker and context
Each workflow can route work across Claude Code, Codex CLI, Cursor, Gemini, browser sessions, MCP tools, provider APIs, and lower-cost models. Goals, schedules, secrets, skills, and company knowledge stay attached to the workflow instead of disappearing across terminals and chat sessions.
Measure the business outcome
Every run keeps logs, reports, screenshots, browser state, artifacts, cost, approvals, and goal evidence in one operating record. Pulse turns that record into health signals and questions that need a human decision.
Improve with bounded changes
Auto Improve studies evidence from previous runs and proposes changes to instructions, model routing, skills, schedules, and workflow configuration. Risky writes, production actions, and major replans remain behind explicit approval.