Deploy AI apps on your own server

Push your AI app's code to git. tidoly builds it and runs it on a server you own — with its database, cache and API keys wired in.

Can I deploy my own LLM app?

Yes. Point tidoly at a git repository. You do not need a Dockerfile: the build uses Cloud Native Buildpacks, and if your repository has a Dockerfile, that is used instead. Every push to the branch you track builds and deploys again.

Where do my API keys go?

Upload your .env file. Values that look like secrets are stored as Kubernetes Secrets, and the rest become plain settings. You see a preview first and choose what to keep.

Does it come with a database and a cache?

Yes. Choose "new database" and tidoly creates a PostgreSQL database for the app and injects the connection values. Valkey, a Redis-compatible cache, installs from the catalog and its address arrives in your environment variables.

Can an AI agent do the deploy?

Yes. The control plane runs an MCP server. With a token you issue, an agent can create the app and its git repository, build, deploy, publish it at an address, read logs and set up sign-in.

• The token belongs to one project, expires, and can be revoked at any time.
• An agent cannot delete apps or installs.
• Every call is recorded, and each app shows its own history.
• The agent runs on your laptop and reaches the server over the site's private network (VPN). The MCP endpoint is not on the public internet.

What is not supported yet?

GPU scheduling and model serving. tidoly runs your app code and the services around it; it does not host models. It also runs on a single server today.

Related

• PostgreSQL for your app
• Redis and Valkey for your app
• Keycloak and OIDC sign-in
• Deploy without YAML
• Self-hosted PaaS compared

Start free — a trial, no card required.