Skip to main content
Daytona runs isolated sandboxes (containers, VMs and macOS) for executing AI-generated code, managed through its API, SDKs and dashboard. The following table shows what CostGraph runs on this provider.

​
Connect

Connect Daytona on Integrations, and CostGraph starts sandboxes in your Daytona account. You don’t run anything yourself.
  1. In CostGraph, go to Integrations > Sandbox > Daytona.
  2. Enter these fields:
    • API key (DAYTONA_API_KEY, required): A Daytona organization API key from the Keys page of the Daytona dashboard.
  3. Click Connect.
For CI runners, CostGraph sets CPU and memory unless providerTemplate names a snapshot, which keeps its own size. Auto-stop is off and the sandbox is deleted at the job timeout.

​
Your own dispatcher

To keep the key out of CostGraph or reach a private network, run your own dispatcher instead. Set provider to daytona and the preceding fields as environment variables, as described in Run your own dispatcher. To see what Daytona costs, connect the Daytona integration.

​
Key and permissions

DAYTONA_API_KEY is a Daytona organization API key, created in the Daytona Dashboard at app.daytona.io/dashboard/keys (Create Key) or via POST /api/api-keys. Scoped. Minimum: write:sandboxes (create) and delete:sandboxes (delete). Read/list of sandboxes is always granted, and running commands (toolbox: processes, files, terminal) is not a grantable scope: any valid key for the organization can exec in its running sandboxes. Note write:sandboxes also allows mounting any org secret and reading usage/spend. Sources: www.daytona.io: api keys, www.daytona.io: sandboxes, www.daytona.io: limits, app.daytona.io: keys.

​
Limits

  • Auto-stop defaults to 15 minutes and “triggers even if there are internal processes running in the sandbox”; set auto-stop to 0 to run indefinitely. VM sandboxes default to 60-minute auto-pause instead when neither is set. (source)
  • Default 1 vCPU / 1 GB / 3 GiB disk; per-sandbox max 4 vCPU / 8 GB / 10 GB disk. (source)
  • Org tier caps (vCPU / RAM GiB / storage GiB): Tier 1 10/10/30, Tier 2 100/200/300, Tier 3 250/500/2000, Tier 4 500/1000/5000; limits apply to the org’s default region. (source)