Tutorial · Providers
How to connect your own LLM provider
This tutorial shows you how to bring your own LLM to VegaDūta. You paste an API key from any of 13 supported providers — Anthropic, OpenAI, Azure OpenAI, Google Vertex, Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok, OpenRouter, Ollama, or Sarvam — the platform validates it live, stores it encrypted, and loads that provider's current model catalog so you can choose models per agent.
To be clear about when your keys are used: the free tier runs on the platform-hosted model pool, and your own keys take over on paid tiers.
Step by step
Open Providers in the console
Sign in at app.vegaduta.ai and open the Providers screen. This is where every model credential in your workspace lives — one place to add, replace, or remove keys.
Pick your provider
Choose from the 13 supported providers: Anthropic, OpenAI, Azure OpenAI, Google Vertex, Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok, OpenRouter, Ollama, and Sarvam. Ollama covers self-hosted models; Sarvam covers Indian-language models.
Paste the API key
Paste the key from your provider's dashboard. VegaDūta validates it live against the provider before saving — a bad or expired key fails right here, not later in a customer conversation. Valid keys are stored encrypted.
Let the model catalog load
Once the key validates, the platform loads the provider's current model list into your catalog. You see what your key can actually reach — no hand-maintained model lists to keep in sync.
Select models per agent
Open an agent and pick its model from the catalog. Different agents can run on different providers — a fast cheap model for triage, a stronger one for complex work — within the same workspace.
Optionally set budgets and fallback
You can set per-user daily budgets to cap spend, and configure fallback so that if a provider fails, agents continue on an alternative instead of going down with it. Both are optional but recommended before production traffic.
Good to know
A few honest details about how provider keys behave on the platform.
- Free tier requests run on the platform-hosted pool; your own keys are used on paid tiers.
- Keys are encrypted at rest and can be replaced or deleted from the Providers screen at any time.
- Validation happens at save time, so provider-side revocations surface as errors on the Providers screen rather than silently.
Frequently asked questions
Which LLM providers can I connect to VegaDūta?
Thirteen: Anthropic, OpenAI, Azure OpenAI, Google Vertex, Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok, OpenRouter, Ollama, and Sarvam. That spans hosted frontier models, aggregators, self-hosted models via Ollama, and Indian-language models via Sarvam.
Is my API key stored securely?
Keys are validated live against the provider when you save them and stored encrypted at rest. You can replace or delete a key from the Providers screen at any time, and a key that stops working surfaces as an error there rather than failing silently.
Do I need my own API key on the free tier?
No. The free tier runs on the platform-hosted model pool, so agents work without any provider setup. Your own keys come into play on paid tiers, where they are used for your agents' requests.
Can different agents use different models?
Yes. Model selection is per agent: once a provider's catalog is loaded, each agent picks its own model. You can also set per-user daily budgets and a fallback chain so a provider outage does not take your agents down.
See it working in two minutes
The sandbox provisions a real tenant — describe an agent in one sentence and test it, no account, no card. Or browse ~90 industry workflow recipes to see what teams build.
