AI Provider
AI provider prerequisites for Scrydon features
Scrydon ships no models of its own. Every AI feature — Cortex chat, agentic workflows, the analytics assistant, and knowledge extraction — runs against an external LLM, so you must connect at least one AI provider before any AI-backed feature works. Until an AI integration is enabled, the model picker is empty and those features return a "no provider configured" error.
Bring an account and API credentials with one of the supported providers:
| Provider | Notes |
|---|---|
| OpenAI, Anthropic, Azure OpenAI, Amazon Bedrock, Mistral | Hosted APIs. Supply an API key, or cloud credential for Azure/Bedrock. |
| Ollama, self-hosted vLLM | Run models inside your own cluster. Needs GPU nodes; see Resources. |
There is nothing to configure at install time. Once the platform is up, an org admin enables a provider under Settings → Integrations. See the integration catalog for the full list of vendors and models. A search integration such as Exa, Linkup, or Perplexity is also worth adding if you want agents to look things up.
Egress: hosted providers are reached over the public internet. On an egress-restricted or air-gapped cluster, allowlist the provider's API host or run a local model. See Air-Gapped Deployment.