Skip to main content
By the end of this you’ll have:
  • The veri CLI installed locally.
  • An authenticated session.
When you’re ready to spend real GPU time, jump to the workflow docs at the bottom of this page.

​
Install

The CLI ships the Python SDK and CLI together.
The mcp extra adds veri mcp serve for MCP clients such as Claude Code and Cursor. To upgrade an existing install, or add the extra to one, run uv tool install --upgrade 'veri-sdk[mcp]' (pipx: the line above; pip: pip install -U 'veri-sdk[mcp]'). uv tool upgrade and pipx upgrade never add an extra. Verify:
The CLI keeps credentials in ~/.config/veri/config.toml on Linux and ~/Library/Application Support/veri/config.toml on macOS (or $XDG_CONFIG_HOME/veri/config.toml when that is set). veri whoami prints the path in use.

​
Log in

Create an API key from the Veri dashboard, then:
Confirm:
Output (rendered as a table):
Prefer environment variables? Set VERI_API_KEY=vk_... and any CLI call will pick it up. Env vars take precedence over the config file.

​
Where to next

Pick the workflow you came for. Each guide covers the path against https://api.veri.studio and may spend GPU credit.

Train a model

Configure and run hosted training jobs

Deploy a model

Serve base or trained models behind an endpoint

Follow a guided demo

Step-by-step walkthroughs from first run to live endpoint

​
Stuck? Book a 15-minute call

A 15-minute call gets you unblocked faster than a docs page. Tell us what broke and we’ll fix it.

Book a 15-min beta call

Pick any time. If you’re hitting a specific issue, share your account email and any failing job IDs.

​
Spot a bug?

Email daniel@veri.studio with what you expected, what happened, and any failing job IDs.