Baseten Grid

Baseten Grid connects your organization’s coding agents to models on Baseten and connected providers. Admins choose each team’s models and budgets; developers use the team’s routes in Claude Code, Codex, OpenCode, or Pi.

​

Get started

For IT and platform teams distributing Grid across machines, see Roll out Grid to developers.

​

Where requests go

Setup changes the coding agent’s model connection. The developer then uses the agent as usual, and its model requests go to Grid with a Grid API key. On a direct route, Grid sends every request to the configured model. On an auto route, a managed classifier chooses among the route’s allowed models and uses its default route when it makes no selection. Baseten bills Model API usage. Connected providers bill usage to the account behind the team’s provider key. Developers use their Grid API keys to call Grid; an admin manages the provider keys in Connections.

​

How Grid works

1. Choose each team’s models. Add open-weight models from Model APIs, frontier models from Anthropic, OpenAI, or xAI through your own API key, or both. Add an auto route when you want a classifier to choose among several models for each request. Developers select direct and auto routes from the same model picker.
2. Set spend limits. Set a monthly budget for each member and for the team as a whole, and raise one developer’s limit when they need more. Grid meters spend every 15 minutes, so spend can exceed a limit before new requests are blocked. For more information, see How spend limits work.
3. Let developers keep their workflow. Each developer runs baseten grid setup once. Their coding agent lists the team’s models and keeps its shortcuts, MCP servers, and plugins.
4. Track spend and usage. See spend, tokens, and cache hit rate by team, developer, model, and provider, and adjust models and limits as your teams’ needs change.

​

Key terms

• Route: a named entry a coding agent can call. A direct route calls one model; an auto route chooses among several direct routes.
• Connection: provider credentials stored on a team, so the team’s routes can call that provider’s models.
• Harness defaults: the model each coding agent uses for its main work and for background tasks when a developer runs setup.
• Spend limit: a monthly cap, in USD, on a developer’s or a team’s Grid spend. Limits reset at the start of each UTC calendar month.
• Grid API key: the key baseten grid setup creates for coding agents on one machine. It’s separate from a developer’s personal Baseten API key, and spend limits apply to it.

​

When to use Grid

Grid is for organizations that want central control over coding agent models and spend. For other cases:

• To connect your own coding agent to Model APIs without admin controls, use Baseten Switch or a direct provider setup.
• To give customers or applications access to models across providers with federated API keys, use Frontier Gateway.