Your AI agent, managed.
Centaur is an open-source AI agent that lives in your Slack. It writes code, browses the web, queries databases, and connects to 1,000+ external services via tool plugins. This dashboard gives you a managed instance on Hetzner Cloud—no Kubernetes required.
60s to deploy
1,000+ tool integrations
7× cheaper than AWS
MIT licensed
What is Centaur?
Centaur is an open-source AI agent built by Paradigm. It connects to Slack, executes tasks using LLMs, and orchestrates tool calls across code execution, web browsing, file management, and external services via Composio.
This dashboard is Centaur-as-a-Service: managed deployments so you don’t have to run Kubernetes yourself. Create an instance, bring your own inference keys, and the provisioner handles the rest—Terraform, Helm, TLS, network isolation.
What you get
- Isolated namespace
- Test deployments run as separate Kubernetes namespaces on shared infrastructure, with their own Postgres, secrets, ingress, and NetworkPolicy boundary.
- Credential injection
- Add your OpenAI key, Slack bot token, and signing secret through the dashboard. They’re pushed to Kubernetes as sealed secrets—you never touch kubectl.
- Auto-provisioning
- Terraform creates the Hetzner server, k3s bootstraps the cluster, Helm installs Centaur, and cert-manager provisions TLS. The whole pipeline runs in under 90 seconds.
- Slack integration
- Mention the bot in any channel. It runs in a sandboxed environment with your tools, repos, and credentials—fully isolated from every other tenant.
- Composio tools
- 1,000+ external service integrations via Composio. Connect GitHub, Linear, Notion, Jira, and more—the agent uses them as native tools.
- Open source
- MIT licensed. Read the provisioner, audit the Helm chart, fork the dashboard. The entire stack is on GitHub.
How it works
- 1
Create a deployment
Name your instance and choose a Hetzner region. The dashboard creates the record and hands off to the provisioner.
- 2
Add your credentials
Paste your OpenAI API key, Slack bot token, and signing secret into the setup wizard. The provisioner pushes them to your cluster and restarts the agent automatically.
- 3
Talk to your agent
Mention the bot in Slack. It processes your request using your configured model provider, executes tool calls in a sandboxed environment, and replies in the thread. Every invocation is logged to the usage dashboard.
Under the hood
Infrastructure
Hetzner Cloud servers running k3s. OpenTofu for provisioning, Helm for deployment, cert-manager for TLS. NetworkPolicy isolates each tenant.
Dashboard
Next.js 16 with Clerk auth, Drizzle ORM, and Railway Postgres. Manages deployments, credentials, and usage tracking.
Agent runtime
Centaur core with Codex integration. Built-in tools for code execution, web browsing, file management, and database access—plus 1,000+ external services via Composio.
Beta access
Hosted deployments are free during the current test window. Bring your own inference keys; billing and subscriptions are disabled for now.
Pricing
Free while the hosted beta is open. Bring your own inference keys for OpenAI, Anthropic, or whichever model provider you want to use.
Centaur Instance
Free
Test deployments run on shared infrastructure for now. Add your own inference keys after provisioning; Centaur does not charge for usage during this beta window.
Ready to deploy?
Sign up, create a deployment, and have a running Centaur instance in under two minutes.