Start building
or drag nodes from the left — or start from a template:
Drag. Connect. Deploy GPU workflows.
New tenants start with $25 of free compute credit.
Rent a whole GPU for your own models — billed per hour. Use it inside workflows (pick it in a Model node) or call it directly via an OpenAI-compatible endpoint.
Run a saved workflow automatically on a schedule — each fire is billed like a normal run.
Use these for programmatic access — pass X-API-Key.
Manage your profile, organization, and password. Changes apply to your EntBox AI account.
Restore an earlier saved version — restoring snapshots the current one first, so nothing is lost.
Describe a new workflow — or, with a flow already on the canvas, describe a change to it. The local model uses EntBox AI's own nodes.
The exact text chunks indexed for retrieval — decrypted just for you.
POST to this URL to run “” from anywhere. The request body (text/JSON) is injected into the workflow's first Dataset node.
Approve this output?
Add your own Claude or OpenAI key. After a race, that frontier model grades each local model's answer for accuracy and badges the most accurate. Your key is encrypted per-tenant and used only for judging — it never appears in responses.
Generate a question/answer test set from this bucket, then grade retrieval & answers with your frontier judge key. Metrics follow RAGAS: Faithfulness, Answer Relevancy, Context Precision, Context Recall. Use with public / non-sensitive data — questions and chunks are sent to your judge provider.
Simulated payment — instantly credits your account.
Tell us what went wrong. We'll automatically attach a snapshot of your recent runs, their logs, your environment, and live GPU/fleet health from around the time of the problem — so we can investigate without asking you to reproduce it.
Fault reports you've filed and their status. We'll also drop replies from support into your 🔔 notifications.