AI · Models
LLM Connectors on VegaDūta
VegaDūta is model-agnostic: you bring your own API key for any of 13 LLM providers and every agent on your tenant can use them. Supported providers are Anthropic (Claude), OpenAI, Azure OpenAI, Google Vertex AI, Google Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok (xAI), OpenRouter, Ollama for self-hosted local models, and Sarvam for Indian-language models.
The platform fetches each provider's live model catalog, validates your key when you add it, enforces per-user daily budgets, and falls back automatically to a platform-hosted model pool when needed.
Thirteen providers, one interface
Add a key once and the provider's models appear across the platform — agent configuration, workflows, and the playground. Model lists are fetched live from each provider rather than hard-coded, so new models show up without waiting for a platform update.
- Hosted APIs: Anthropic, OpenAI, Azure OpenAI, Google Vertex AI, Google Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok (xAI), OpenRouter
- Self-hosted: Ollama, for running open models on your own hardware
- Indian languages: Sarvam, alongside its speech and translation suite
Keys are validated and spend is bounded
Keys are checked against the provider when you save them, so a typo fails immediately instead of at 2 a.m. in a customer conversation. Per-user daily budgets cap what any one user can spend, and encrypted storage keeps credentials out of logs and prompts.
Fallback and the free tier
On the free tier, agent traffic routes through a platform-hosted model pool — you can start without any API key at all. Paid tiers run on your own keys, and an automatic fallback chain can route to the platform pool if a provider call fails. Plan details live at /billing.
Frequently asked questions
Which LLM providers does VegaDūta support?
Thirteen: Anthropic (Claude), OpenAI, Azure OpenAI, Google Vertex AI, Google Gemini, AWS Bedrock, Mistral, DeepSeek, Groq, Grok (xAI), OpenRouter, Ollama, and Sarvam. Each is bring-your-own-key, and Ollama lets you point at a self-hosted local model server.
Do I need my own API key to use VegaDūta?
Not to start. The free tier routes through a platform-hosted model pool, so agents work with no key at all. On paid tiers you connect your own provider keys and your traffic runs on them.
Can I run local or self-hosted models?
Yes, through the Ollama connector — point it at your own Ollama server and its models become selectable like any hosted provider. For fully in-browser models, see the Edge AI capability, which runs models on-device via WebLLM.
What happens if my provider goes down or my key hits a limit?
An automatic fallback chain can route the call to the platform-hosted model pool so the conversation continues. Per-user daily budgets also cap spend, so one runaway user cannot exhaust your provider account.
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.
