Skip to main content
Environments keep resources apart inside one workspace, such as staging and production. Each has its own instances, volumes, API keys, firewall and monthly budget. When you connect your agent over MCP, you pick one environment on the sign-in page. The agent works only there: it sees and launches resources in that environment, and that environment’s budget caps what it can spend. It can’t create, delete or switch environments, or change a budget. You do that in the dashboard.

”Which environment are you working in, and how much budget is left?”

  1. list_environments: the environment this connection works in
  2. get_environment_budget: its cap, what it spent this month and what is left
I’m working in staging, the environment you picked when you connected me. Its budget is $500 this month; $212.40 is spent and $287.60 is left.

”What’s running here?”

  1. list_instances: the instances in this environment
  2. list_volumes: the volumes in this environment
In staging:
  • staging-inference (A100): running
  • staging-models (200 GB volume)
The agent only sees the environment it is connected to. To see every environment at once, open the dashboard.

”Set up a production environment for our ML pipeline.”

The agent can’t create environments, so it tells you how:
I can only work in staging, so I can’t create environments. A workspace owner or admin can add production in Settings, Environments, and give it a monthly budget. To have me work there, disconnect Runcrate in your client and sign in again, picking production.
To move your agent to another environment:
1

Create or pick the environment

In the dashboard, open Settings, Environments. Create one if it doesn’t exist and set its monthly budget.
2

Disconnect Runcrate in your client

For example /mcp in Claude Code, or remove the connector in Claude or ChatGPT.
3

Connect again

Sign in and pick the new environment on the consent page.
Want separate agents for staging and production? Sign each client in to its own environment.

”We’re done testing. Clean up staging.”

The agent can clean up what is in its own environment, and asks before anything destructive:
  1. list_instances and list_volumes: what is still in staging
  2. delete_instance: terminates staging-inference, and its billing stops
  3. delete_volume: deletes staging-models
Terminated staging-inference and deleted staging-models (200 GB). Staging is now empty. To remove the environment itself, delete it in Settings, Environments. The default environment can’t be deleted.

Tools used in this workflow