The AI command center
for your company
Build internal agents like software. Define their skills, tools, persona, and guardrails as code, then deploy them wherever your team works.
How a company runs on Kortix.
Your whole company as code — on the agents and the API keys you already have, multi-tenant for the whole org, hosted wherever you want.
Create
Every project is just a Git repo
kortix init scaffolds a directory. That directory is the product.
- One kortix.toml, statically parsed for config — validate it like code
- .opencode/ holds your agents, skills, commands, hooks — all just files
- Add any other files you want; the repo is the single source of truth
- Secrets never touch the repo — they live scoped in the cloud
Build
Coding agents, doing your company's knowledge work
The harnesses that changed how we code — OpenCode, Claude Code, Codex — now run the rest of the work.
- Agents, skills, and tools from the ecosystem you already use — nothing new to learn
- kortix dev runs OpenCode locally against your config to test everything
- Edit as files in your IDE, in the web UI, or let an agent edit itself
- Technical, non-technical, or the AI — anyone improves the system in one place
Connect
Your keys. Your models. One token.
Bring your own API keys — or your existing ChatGPT subscription. The gateway fronts every provider.
- Provider- and model-agnostic: plug in your own keys, subscriptions, even your own LLMs
- LLM gateway = one token for every model, with logs, observability, cost tracking
- 3,000+ one-click integrations — or any MCP / HTTP / OpenAPI / GraphQL connector
- Sandboxes never see a single secret; agents only hold one scoped Kortix token
Models
Bring any provider — routed per session, keys stay in Secrets
Ship
One deployment. The whole org.
kortix ship is the entire deploy — and it's multi-tenant from day one, not a single-tenant toy.
- First run creates the cloud project; every run after just pushes changes
- Multi-tenant by design: hundreds of users, everyone running their own agents
- Invite anyone with kortix access — to them it all feels like ChatGPT
- Live on web, desktop, mobile, and Slack the moment it ships
Run
1 session = 1 sandbox = 1 branch
Every session is an isolated computer on its own branch — user-scoped, down to the token.
- Full isolation: run thousands of agents in parallel, nothing collides
- One user-scoped Kortix token per sandbox — never a pile of API keys
- Finished work returns as change requests: review, then kortix cr merge
- Main stays the truth; every agent action is auditable Git history
Own
Open source. Self-host in one command.
Managed cloud, your VPC, on-prem, even air-gapped — the same product everywhere.
- kortix self-host init runs the full platform from Docker images
- Run on your own infra with your own models — nothing phones home
- You own everything: the data, the agents, the skills, the repo itself
- Switch hosts anytime with kortix hosts — no lock-in, by design
For everyone
One product. One Git repo.
Just ask. Work comes back done.
No prompts to master, no tools to learn. Anyone on the team describes what they need in plain language — and the same agents that build software deliver it, end to end, wherever you work.
Built on files you can actually open.
No SDK to learn, no API maze. Your agents, skills, and tools are plain files in a folder — clone the project, change what you want in your own editor, and ship it back.
Enterprise & security
Enterprise-ready. Hosted on your terms.
The architecture is the security model — and you can run all of it on infrastructure you control.
Run your company on AI.
Free to self-host. Managed cloud from $20 / seat + usage. Spin up your first agent today — or have us map it to your workflows in a live demo.
Open source · SSO, RBAC & on-prem · No lock-in