Providers
Providers
Neutron AI works beside your model provider. Your application requests scoped memory from Neutron, builds a provider request with the SDK, and sends that request with your own provider credentials.
Supported Provider Guides
| Provider | Adapter ID | Request family | Remote MCP |
|---|---|---|---|
| OpenAI | openai | Responses API | Native |
| Anthropic Claude | anthropic | Messages API | Native |
| Google Gemini | google_gemini | generateContent | Context injection |
| Mistral AI | mistral | OpenAI-compatible chat | Local tools |
| xAI | xai | OpenAI-compatible chat | Local tools |
| DeepSeek | deepseek | OpenAI-compatible chat | Local tools |
| Groq | groq | OpenAI-compatible chat | Documented by provider |
| Cohere | cohere | Chat API | Local tools |
| Perplexity Sonar | perplexity | OpenAI-compatible chat | Local tools |
| OpenRouter | openrouter | OpenAI-compatible chat | Local tools |
| Azure OpenAI | azure_openai | Responses-compatible | Verify tenant policy |
| Amazon Bedrock | amazon_bedrock | Provider messages | Local tools |
| Google Vertex AI | vertex_ai | Gemini-style generateContent | Context injection |
Shared Setup
npm install @neutron-ai/sdk
export NEUTRON_API_URL="https://neutronai.dev/api"
export NEUTRON_API_TOKEN="<server-issued-token>"
export NEUTRON_TENANT_ID="<tenant-id>"
export NEUTRON_SCOPE_ID="<authorized-scope-id>"
import { NeutronAIClient, createNeutronConnector } from "@neutron-ai/sdk";
const client = new NeutronAIClient({
baseUrl: process.env.NEUTRON_API_URL!,
token: process.env.NEUTRON_API_TOKEN!,
tenantId: process.env.NEUTRON_TENANT_ID!
});
const connector = createNeutronConnector(client, {
scopeIds: [process.env.NEUTRON_SCOPE_ID!],
agentId: "agent_support",
cachePolicy: {
mode: "prefer_cache",
keyMode: "intent_profile",
ttlSeconds: 300
}
});
Keep provider API keys in your application. Neutron AI does not need them.