Introducing Tanvrit Compute
A distributed compute fabric for CI, AI/ML inference, and batch workloads.
Today we're launching Tanvrit Compute — a distributed compute fabric that runs CI jobs, AI/ML inference, and batch workloads on machines you own, machines you rent, or machines in our shared pool.
Why we built it
Every team eventually needs a second compute surface: - A CI runner fleet that can scale past the free minutes on GitHub Actions. - GPU-backed inference nodes for production ML. - A cheap, fast place to run nightly batches that don't belong on a pager-duty cluster.
The usual answer is a one-off stack of Terraform modules, a half-maintained Jenkins, and enough YAML to bury a small engineer. Tanvrit Compute is the opposite: one SDK, one portal, one CLI, and a single agent binary that runs on every OS.
What's in the box
- Agent — a ~25 MB GraalVM-native binary. Installs with one shell command,
- Portal (this site, under /app/) — submit jobs, watch logs tail, browse nodes,
- CLI —
tanvritcommand with login, submit, status, logs, cancel, retry, - Docker, GitHub Actions, Services — submit a raw command, a Docker image with
Who it's for
If you're a one-person team paying for GitHub Actions minutes out of pocket, or a 100-engineer org that wants their M4 Pro desktop and a few rented H100s to speak the same job API, this is for you. Start at compute.tanvrit.com.
The free tier includes $30/mo of credits. No card required to start.