Agenta
A self-hosted AI agent workspace with Ollama support, shared files, and approval controls. Run it on your servers or use the hosted cloud service.
Agenta is an MIT-licensed, open source workspace for teams that want AI agents to handle tasks in chat and continue recurring work in the background. You can self-host it with Docker Compose or Helm to keep agents and workspace data on your infrastructure, or use Agenta Cloud as a hosted service.
Agents can plan work, use connected tools, and adapt their approach rather than follow a fixed sequence of workflow steps. You build them through conversation and refine their work with feedback. Templates cover jobs such as code review and customer support, while a shared browser and file workspace gives agents somewhere to research, write documents, and save results.
Model choice stays flexible. Agenta supports self-hosted models through Ollama and cloud providers including OpenRouter and Gemini. Self-hosted deployments can also use existing Claude or ChatGPT subscriptions. Claude Code, Pi, and Codex provide the agent execution layer, and you can switch models or execution tools without rebuilding the agent. Using a cloud model still sends requests to that provider.
Teams can share agents with role-based access and work with them in Slack or Telegram. MCP connections link agents to apps such as GitHub, Notion, and HubSpot; reusable skills carry task instructions between agents. Scheduled runs and events in connected apps can start background work.
Tool permissions let you allow, block, or require approval for actions. Agenta records model and tool calls, tracks usage and estimated costs, and keeps configuration history so teams can investigate failures and compare agent changes.
