Deploy Kortix — Self-Hosted AI Agents

Deploy and Host Kortix — Self-Hosted AI Agents for Your Team

Kortix is an open source platform where AI agents do real work for your whole team. You give an agent a goal, such as "research our top competitors and draft a summary", "fix this bug and send me the change" or "every Monday, pull last week's numbers into a report", and it plans, uses your tools, writes and runs code, and keeps working after you close the tab. A person reviews and approves what it changes.

This template gives you your own private Kortix: the full app on your Railway project, with your data in your database and your own AI model keys.

About Hosting Kortix

One click sets up everything Kortix needs: the web app, the API, the LLM gateway and its database, sign-in and file storage (8 services, wired together with HTTPS domains). You don't need a server, a domain or a terminal.

What you get once it's running:

• AI agents for the whole team. Everyone shares the same agents, skills, company knowledge and tool connections.
• Work that keeps going. Agents run in the background, 24/7, and can start on a schedule or when something happens in your tools.
• 3,000+ app connections (with a Composio key), plus MCP, OpenAPI and any HTTP API.
• Any model, your own keys: Anthropic, OpenAI, OpenRouter and more, at your provider's prices.
• Everything as code. Your agents, skills and rules live in a GitHub repo you own, versioned like software.
• Open source: read it, change it, fork it.

Each agent session runs in its own isolated cloud computer on Platinum, a sandbox service for AI agents. That's why the template asks for a Platinum API key.

Common Use Cases

• A team AI assistant that works across your tools, not just a chat box
• Research and reports: market research, competitor tracking, weekly summaries
• Coding agents that fix bugs and write features, sending each change for review
• Recurring operations: scheduled jobs and reactions to events in your apps
• Trying an open source, self-hosted AI agent platform before rolling it out to your company

Dependencies for Kortix Hosting

• A paid Railway plan (Hobby or higher): the template runs 8 services, and Railway's Free plan stops at 3.
• A Platinum API key (required): agents run in Platinum sandboxes. Create one at platinum.dev (prepaid credits, billed per second while agents work).
• The email of the first admin (required).
• A GitHub account or organization: Kortix keeps every project in a repo.
• A key from an AI model provider, added in the app after deploying.
• Optional: an SMTP URL for email confirmation and magic links, and a Composio key for the app connector catalog.

Deployment Dependencies

• Kortix
• Self-hosting docs
• Kortix on GitHub
• Platinum

After the deploy (about 10 minutes)

1. Open the Kortix service's domain and sign up with the admin email right away.
2. Signed in as the admin, open /admin/git and connect GitHub. Install the GitHub App on an organization so Kortix can create project repos. On a personal account, create each repo on GitHub first, then import it.
3. Add your model key in Settings → Model.
4. Create a project and give an agent its first task.
5. To stop others from signing up, set DISABLE_SIGNUP=true on the Auth service.

Why Deploy Kortix on Railway?

Self-hosting Kortix normally means renting a server, pointing a domain at it and running an installer in a terminal. On Railway it's one click: Railway runs the services, gives them HTTPS addresses, and keeps your data on volumes in your project. You get your own instance, with your data and your keys, without managing a server. Update to a new Kortix version by redeploying.