Open source AI assistant that executes complex tasks from market research to travel planning through natural conversation, with full transparency and community-driven development.

At a Glance:

Kortix is an open-source, self-hostable AI command center that orchestrates a workforce of specialist agents through a single config repository, running isolated sandbox sessions and producing approved change requests to deliver real output like code, reports, and deployments.

Overview:

Kortix is an open-source platform for building and managing a company’s entire AI-powered workforce from a single, version-controlled git repository. It functions as a command center where specialist agents, reusable skills, 3,000+ app connectors, secrets, and automated triggers are defined as code. Instead of providing a chat interface, it deploys agents into disposable, isolated Linux sandboxes that run on their own branches and produce concrete deliverables. Any work an agent completes must be submitted as a change request and approved before reaching the main branch, forming a reviewable, self-improving process that runs on-demand, with human assistance, or fully automated.

Key Decision Points:

  • Deployment model: It is open-source and designed for self-hosting on a laptop, VPS, VPC, or air-gapped infrastructure, with a managed cloud option also available.

  • Execution model: Agents operate in disposable, isolated Linux sandboxes with their own branches; only explicitly committed and approved changes survive through a change request mechanism.

  • Pluggable AI models: The platform is designed to work with any model provider using your own keys or existing subscriptions to services like ChatGPT, Claude, or Cursor.

  • Integration surface: It includes one-click connectors for over 3,000 apps and supports MCP, OpenAPI, GraphQL, and raw HTTP connections, all brokered through a single server-side token.

  • Agent identity and scope: Agents are defined as Markdown personas with scoped access to tools, can be installed with a click, and are capable of rewriting themselves.

Core Features:

  • Code-defined workforce: Agents, skills, connectors, automations, and memory are all defined in one version-controlled, diffable config repository.

  • Isolated sandbox execution: Every agent session runs in its own disposable Linux environment on a dedicated branch, preventing interference and making only approved commits permanent.

  • Change request workflow: All agent-generated work must pass through a human-approved change request to reach the main branch, creating a reviewable audit trail.

  • Automated triggers: Sessions can be initiated on-demand, via cron schedules, or through signed webhooks for event-driven automation.

  • Multi-agent parallelism: Thousands of agents can run concurrently on the same config, each fully isolated and contributing work back through change requests.

  • Managed secrets and memory: Secrets are encrypted, scoped, and injected at runtime without exposure to the model, while a persistent memory system compounds company knowledge over time.

Use Cases:

  • Company operators can version and govern their entire AI workforce, automations, and learned facts as a reviewable codebase.

  • Developers can deploy parallel, isolated agent sessions that produce concrete deliverables like reports, code, and deployments on a schedule or trigger.

  • Self-hosters and security-conscious teams can run the entire platform on air-gapped or VPC infrastructure, using their own model provider keys.

Open-Source Alternative Value:

Kortix provides an open-source, self-hostable alternative to proprietary AI agent orchestration platforms, with the entire company configuration—agents, skills, integrations, and memory—stored as plain code in a git repository. This approach enables portability, auditability via standard tools like grep, and operation fully air-gapped. The agent execution model, based on isolated Linux sandboxes and human-approved change requests, offers a transparent and governable mechanism for producing real deliverables without relying on a black-box managed service.

ShareXLinkedInReddit

Related tools

Project stats

Stars

19,866

Forks

3,420

License

Apache-2.0

Metadata

Alternative to
Manus