Sub-processors

Metaphi, Inc. uses the following sub-processors to run its services, including Humboldt and the Humboldt Slack app. Each one processes customer data only to provide the function listed, under a written agreement, and is bound by the privacy policy and terms. This list is updated before a new sub-processor handles customer data; write to agent@metaphi.ai to be notified of changes.

Infrastructure

Sub-processor	Purpose	Location
Modal Labs, Inc.	Compute for the Humboldt server and per-session sandboxes; encrypted volumes for the application database and session records	United States
Google LLC (Google Cloud)	Object storage for session artifacts, DNS, and model inference through Vertex AI	United States
Google LLC (Firebase Authentication)	Sign-in with Google or GitHub for personal sessions	United States
GitHub, Inc.	Repository access through the Humboldt GitHub App; one-hour tokens, never stored	United States
Vercel Inc.	Hosting of metaphi.ai and humboldt.metaphi.ai web pages	United States

Model providers

Prompts and outputs are sent as stateless API calls. Under each provider's API terms, customer content is not used to train the provider's models.

Sub-processor	Purpose	Location
Anthropic, PBC	Claude models for inference	United States
OpenAI, L.L.C.	GPT models for inference	United States
OpenRouter, Inc.	Routing to additional inference providers, including Kimi models; prompt logging disabled	United States

Communication

Sub-processor	Purpose	Location
Slack Technologies, LLC	Delivery of messages between a workspace and Humboldt, for workspaces that install the Slack app	United States

Last updated 11 September 2026.