kortix-ai/suna

What it solves

Kortix provides a centralized "command center" for managing a workforce of AI agents that perform actual business tasks rather than just chatting. It solves the problem of AI fragmentation and lack of ownership by treating an entire company's operational logic—agents, skills, memory, and configurations—as a version-controlled git repository that the user owns and can self-host.

How it works

The system operates by treating a company as a codebase. Users define agents and skills in a kortix.yaml config file. When a task is initiated, Kortix creates a session, which is an isolated Linux sandbox running on its own git branch. An agent works within this sandbox, performs actions, and submits its results via a change request (similar to a pull request). The user reviews and merges these changes into the main branch to permanently update the company's state or deliverables.

Who it’s for

It is designed for organizations and teams that want to deploy specialist AI agents to handle complex workflows, automate reports and code, and maintain full auditability and ownership of their AI infrastructure without being locked into a proprietary black-box platform.

Highlights

• Git-based Operations: Every action, agent, and piece of company memory is versioned, diffable, and stored in a git repo.
• Isolated Sandboxes: Each session runs in a disposable microVM/Linux sandbox to ensure security and stability.
• Extensive Connectivity: Supports over 3,000 apps via connectors, as well as MCP, OpenAPI, and GraphQL.
• Flexible Deployment: Can be run on Kortix Cloud or self-hosted on a laptop, VPS, or air-gapped environment.
• Bring Your Own Model: Compatible with various LLM providers and existing subscriptions (ChatGPT, Claude, Cursor).