Skip to main content
Version: v2.0

What is Agenta?

Agenta is an open-source workspace where you build agents that automate and augment your work.

ChatGPT gives you one prebuilt agent that already works for everything. Agenta starts empty. You build an agent for a job, give it the tools and knowledge that job needs, and improve it as you go.

You build an agent by chatting with it. You describe the work, connect the apps it needs, and correct it through feedback.

Two ways to use an agent​

Work with it in chat. The agent is a coworker. You give it a task, it does the work, and you go through the result together.

Run it in the background. The agent starts on a schedule, or when something happens in an app you connected, and works without you.

Features​

Workspaces for you and your agents. Work with your agent on files in a shared workspace. Together you can write documents, organize research, or maintain a wiki.

Human approval and permissions. Choose which actions an agent runs on its own, which need your approval, and which are blocked. Set a default for the agent and a permission on individual tools.

Background agents. Run agents on a schedule, or start them when an event happens in a connected app.

Tracing, usage, and versions. Inspect every model and tool call. Track model requests, token usage, and estimated cost for each agent. Compare versions of an agent's configuration to see what changed.

Team access. Share agents with your team and control access by role.

Integrations. Connect agents to the applications you use, through built-in integrations or through MCP servers.

Why Agenta​

Use your Claude or ChatGPT subscription​

When you self-host Agenta, you can run agents with a Claude or ChatGPT subscription you already pay for, instead of moving every task to metered API billing.

Choose your harness and model​

A harness is the runtime that drives the model: it calls the tools the model asks for and feeds the results back. Agenta supports Claude Code and Pi. You can switch harness or model without rebuilding your agent, and run models from the major providers, from an OpenAI-compatible endpoint, or from a model you host yourself. See Harnesses and models.

Build on open agent standards​

Define your agent with AGENTS.md, skills, and MCP servers. Skills and MCP servers from the wider agent ecosystem work in Agenta.

Where to run it​

Agenta Cloud is the hosted version. You can also self-host Agenta on your own infrastructure.

Agenta is open source under the MIT license. You can self-host it, modify it, and use it in commercial projects.

Start here​