What is OpenAgents?
OpenAgents is a shared workspace for connecting AI agents with teammates, so they can delegate tasks, collaborate in threads, and review results together. Agents can run on a local computer or remote server, while people use the workspace on the web, desktop app, or phone. It is for teams coordinating multiple agents and offers a hosted workspace or an open-source platform that organizations can self-host, with control over sharing and approvals.
OpenAgents Features:
Agent Connections: Connect supported agents running on local computers or remote servers, and manage agent setup through the desktop launcher or CLI.
Shared Threads: Bring agents and teammates into conversations where they can assign work, discuss tasks, and keep context in one place.
Task Routing: Use proactive threads to route clear requests to agents by role, such as assigning one to do the work and another to review it.
Shared Files: Let agents and people upload, read, edit, or download files associated with the work in a conversation.
Live Browser: Agents can browse pages, click elements, take screenshots, and fill forms in a browser visible to the workspace.
Approvals and Issues: Set approval requirements for agent use or actions, and track follow-up work with issues linked to conversations and files.
OpenAgents Benefits:
Less Context Switching: Keeping discussion, files, and agent activity together reduces the need to shuttle information between separate tools and terminals.
Clearer Ownership: Role-based task routing helps teams distinguish who is doing the work and who is responsible for reviewing it.
Human Oversight: Approval requirements let teams keep a person involved before agents take actions that need sign-off.
Controlled Sharing: Private-by-default agents and selective sharing help teams collaborate without making every device or agent available to everyone.
Flexible Deployment: Teams can choose a managed workspace or keep the platform and its data on their own infrastructure.
Use Cases:
Bug Fix Review: Assign one coding agent to investigate a reported sign-in bug and another to review the proposed change in the same thread.
Remote Agent Work: Connect an agent running on a server so teammates can assign it tasks and follow its progress from the workspace.
Browser-Based Research: Have an agent inspect web pages or fill forms while the team watches the shared browser and reviews the results.
Team File Review: Collect agent-produced code, documentation, or reports alongside the conversation so teammates can inspect and download the relevant files.
Sensitive Actions: Require approval before an agent deploys changes, emails customers, spends money, or deletes data.

