Providers & API keys

The Providers & API keys section of Settings is where you connect the LLM providers your workspace uses. Add a provider key here and its models become selectable in chats and graph nodes. Settings are scoped to the active workspace (personal or team) and managed by the ProviderManager component.

Supported providers

Provider	What you provide
OpenAI	API key
Anthropic (Claude)	API key
Google Vertex AI (Claude and Gemini)	A service-account JSON, plus the region
Mistral	API key
Gemini	API key
HuggingFace	API key
NVIDIA Build	API key
DeepSeek	API key
GLM (z.ai)	API key
Kimi (Moonshot)	API key
Novita	API key

One row per provider, each with the shape of key it expects. The sentence at the top names the scope the keys belong to — here a personal account. Vertex is the odd one out at the bottom: a service-account JSON and a region rather than a key. The stored key’s last characters are masked.

Caution

Vertex is the one that is not a key. What you paste is the service account your Google Cloud project uses — the JSON file you download from GCP — and chatty-lab exchanges it for a short-lived token on every call.

One credential, two families. Vertex serves Claude (a partner publisher) and Gemini (Google’s own), and the same service account reaches both — which model goes to which is decided by its id, not by a second credential. The difference that matters to you: each Claude model has to be enabled in your project before it answers, and one you have not enabled shows up as unreachable in the picker — which is the point of the check. Gemini models are Google’s own and need no such step.

Vertex or AI Studio? The same Gemini model is reachable both ways and costs the same. Vertex bills through your Google Cloud project (with its quotas, its IAM and its invoice); AI Studio bills against a plain API key. Configure whichever your organisation already uses — configuring both lists each model twice, once per house.

Note

Local models are configured separately — see Ollama & embeddings for running a local Ollama backend.

Adding a provider key

1. Open Settings and go to Providers & API keys.

2. In the Provider Manager, choose a provider and paste its API key.

3. Save. The provider’s models become available wherever you pick a model — in chats, the model picker, and Studio LLM nodes.

Related settings

Models & providers How models are selected and compared in chats.

Pricing Set per-model token costs so runs and trials report cost.