Agenta

Agenta positions itself as an open-source workspace where teams build AI agents and AI co-workers rather than one-off prompts. You can start from pre-built agent templates, give an agent access to models and tools, and run it from a chat surface, Slack or Telegram so the work happens where your team already communicates. Agents can browse, read and write files, and run scheduled automations that trigger from your connected systems.

What sets Agenta apart is its focus on operational control. Agents connect to more than a thousand integrations and MCP servers, secrets live in a vault the agent can use but not read, and permissions and approval steps decide which actions run automatically and which need a human sign-off. Every run is logged with duration, token usage and cost, so you can trace why an agent produced a certain result and iterate on it with feedback.

For engineering teams, Agenta ships as MIT-licensed software you can self-host with Docker Compose or Helm, keeping agents and data on your own infrastructure with no vendor lock-in. A managed cloud option is available for teams that prefer not to run the stack themselves, and both share the same workspace concepts of agents, skills, sessions and automations.