Bring Your Own LLM Provider

Run Datafold’s AI features on your own LLM inference provider and credentials.

With an LLM Provider integration, your organization brings its own inference provider — your API keys, your endpoints, your cloud account — and Datafold’s AI features run through it.

• Datafold Assistant (in the app and in Slack) runs only on your organization’s own provider. It does not answer until an admin adds one.
• Migrations delivered by Datafold run on Datafold-managed model infrastructure by default. Add a provider if your migration must use your own inference account.

NOTE

To set up an LLM Provider, navigate to Settings → Integrations → LLM Providers and click Add new integration. LLM Provider integrations are managed by organization admins.

​

How it works

• Register a provider with its endpoint and credentials, and select which models to expose. For most providers, Datafold fetches the available model list automatically once credentials are entered.
• Models appear in Datafold’s model pickers, labeled with the integration name.
• Credentials are resolved at request time. Editing an integration’s key or endpoint takes effect on the next request.
• Credentials are encrypted at rest and never returned by the API — secret fields are masked on read and kept unchanged when a masked value is re-submitted.
• Costs are tracked automatically. Datafold tracks token usage and estimated spend per model.

​

Supported providers

OpenAI-compatible (universal)

OpenAI

Anthropic

Google Gemini

Azure AI Foundry

AWS Bedrock

Google Vertex AI

Databricks

Snowflake Cortex

Anything that speaks the OpenAI /chat/completions protocol — Together, Fireworks, Groq, OpenRouter, Mistral, vLLM, LM Studio, Ollama, and most other inference services — works through the universal OpenAI-compatible provider.

​

Common configuration

Every provider shares a few fields:

Field	Description
Name	A display name for the integration (e.g. Azure Prod). Shown next to its models in pickers.
Models	The model ids this integration exposes. Fetched automatically where the provider supports listing; models can always be added by id manually.
Max Output Tokens	Per-response output cap. Pre-filled with a family-appropriate default; override if your deployment differs.
Context Window	Model context size in tokens. Pre-filled with a default; set it to your model’s context size if it differs.
Default model	The model used when this is the default provider and no model is picked. Falls back to the first model when unset.
Default provider	Makes this the organization’s default provider. Only one provider can be the default. A single provider is used automatically without this setting.

​

Model quality matters

Small models, or models without tool calling, perform poorly with the Datafold Migration Agent. We recommend frontier-class models (Claude Sonnet/Opus, GPT-5 family, Gemini Pro/Flash, Mistral Large).