useagenthq/useagent

What it solves

UseAgent provides an open-source, self-hostable workspace for AI agents like Claude Code, Codex, OpenCode, and Pi. It solves the problem of fragmented agent workflows by providing a unified, secure environment where agents can access tools, repositories, and memory while keeping your data and credentials under your control.

How it works

It acts as a control plane that manages agent sessions through a shared event-based architecture. When you assign a task via Slack, the web app, or an API, the system spins up an isolated Linux sandbox (using providers like Daytona or CubeSandbox) for the agent. Every action, tool call, and file change is logged in a Postgres database, allowing you to audit, replay, or intervene in the agent's work at any time. Sensitive credentials remain on your server and are never exposed to the agent's sandbox.

Who it’s for

It is designed for teams and developers who want to use advanced AI agents for coding, research, or automation but require data privacy, compliance, and the ability to self-host their infrastructure.

Highlights

• Unified Interface: Swap between different AI engines (Claude Code, Codex, etc.) while keeping the same thread history, memory, and file context.
• Full Auditability: Every step is stored as an append-only log in Postgres, making sessions durable and replayable.
• Secure Sandboxing: Agents operate in isolated Linux environments with no direct access to your private API keys.
• Extensible Integrations: Connects with Slack, GitHub, and various productivity tools, with an open framework for adding custom connectors.