kortix-ai/suna

What it solves

Kortix provides an open-source, self-hostable operating system for AI agents, designed to act as a secure, enterprise-grade alternative to closed-source platforms like Claude Cowork or ChatGPT Work. It solves the problem of managing AI agents that need to perform real, complex work—such as coding, report generation, or tool automation—within a secure, version-controlled, and auditable environment.

How it works

Kortix treats an entire company's agent infrastructure as a single git repository. Every agent, skill, memory file, and connector is defined as code. When an agent is triggered, it runs in its own isolated, disposable cloud computer (a sandbox). The agent performs its task on a dedicated branch, and any output (like code changes or reports) is submitted via a change request that a human must review and approve before it is merged into the main codebase.

Who it’s for

It is designed for organizations and teams that need to integrate AI agents into their existing workflows while maintaining strict security, governance, and data ownership. It is ideal for teams that want to move beyond simple chat-based AI to agents that can execute tasks in Slack, Microsoft Teams, or across 3,000+ external applications.

Highlights

• Everything is Code: Agents, skills, and configurations are versioned and reviewable in git.
• Isolated Execution: Each task runs in its own disposable, secure Linux sandbox.
• Human-in-the-loop: Features approval gates and change requests to ensure agents only act with human consent.
• Flexible Deployment: Can be self-hosted on-premise, in a private VPC, or used via the managed Kortix Cloud.
• Extensible: Supports 3,000+ integrations, plus custom connectors via MCP, OpenAPI, GraphQL, and raw HTTP.