Decawork

什麼是 Decawork？

Decawork is a control plane for internal AI agents. It is built for IT teams that need a central way to approve agents created in different tools, set access boundaries, and inspect what those agents do.

The product focuses on three core jobs: knowing which agents are running, controlling what they can reach, and keeping an audit trail of actions and approvals. The site describes a model where teams keep building agents in their preferred tools while IT manages them in one place.

Decawork 能做什麼？

Agent inventory

See a company-wide list of managed agents, including who owns them and whether they are live or paused.

Approval gates

Require approval before an internal agent can be allowed to access systems or take action.

Scoped credentials

Give agents short-lived credentials and exact permissions instead of broad access.

Audit trail

Track what happened, which agent did it, and who approved it.

Access control

Define what each agent can access and do across company systems.

Multi-tool support

Manage agents built in different tools from one control plane.

使用情境

“IT oversight for employee-built agents”

Keep one view of agents created by different teams so IT can understand what is running across the company.

“Security review of agent actions”

Inspect action histories and approvals when teams want visibility into how an agent behaved.

“Access management for internal workflows”

Restrict agent permissions to the exact systems and actions needed for a task.

“Operational control for paused or live agents”

Monitor whether agents are active or paused and maintain ownership information.

“Company-wide agent governance”

Apply a common approval and audit process to agents built on different platforms and frameworks.

常見問題

What does Decawork do?

Decawork gives IT teams one place to approve internal AI agents, control what they can access, and review every action they take.

Who is it for?

The site positions Decawork for IT teams and organizations that want company-level control over agents built by different teams.

Does it work with agents built in different tools?

Yes. The site says agents can be built anywhere, including examples such as ChatGPT, Claude Code, Microsoft Copilot, OpenAI Agents SDK, LangGraph, CrewAI, Gemini, Cursor, and custom stacks.

Does the pricing page show a price?

No. The pricing URL shown in the source returns a not found page and directs visitors back to the home page or to book a live walkthrough.