Maya is an Agent 工作管理台 and multi-agent workbench for managing Agent work like task cards, with Goals, Task queues, Approvals, Artifacts, Runtime status, and execution logs in one place.

Which agents can Maya manage?

Maya is designed to coordinate Codex, Claude, OpenClaw, Hermes, cloud Runtimes, local assistants, and human teammates from one traceable workflow layer.

How does Maya make agent work traceable?

Maya records each Goal, Task, Task Run, Approval, Artifact, Runtime event, blocker, retry, and handoff so teams can locate where an Agent is stuck and trace how work changed over time.

Maya workbench

A control plane for complex work with agents.

Maya is a multi-agent workbench for people who need real work to keep moving after the first prompt. It turns a Goal into planned Tasks, routes those Tasks through cloud or local Runtimes, pauses at Approval points, and keeps every Artifact attached to its source run.

The product is built for operators, founders, content teams, analysts, course creators, and technical leads who already use agents for complex work but need queue visibility, audit trails, and reusable outputs instead of scattered chat threads.

Current development is focused on making the core workflow production-ready: clearer queue observability, safer Artifact delivery, stronger local-assistant support, Lark handoff, and provider expansion.

Read the docs View the changelog

Work starts from Goals

Maya keeps the desired outcome, source context, Task plan, approvals, execution records, and final files connected.

Execution stays observable

Workers report queue position, progress, blockers, stop state, and outputs so users know what is happening without opening logs first.

Cloud and local belong together

Hosted workers handle ordinary jobs; local assistants handle private files, internal networks, and machine-local tools.