Providers
OpenAI and Azure OpenAI
Use the openai adapter for OpenAI Responses API requests. Use azure_openai when your Azure deployment supports an OpenAI-compatible Responses flow.
Build a Provider Request
const { request, contextPack } = await connector.createProviderContextRequest(
"openai",
{
task: "Answer the customer using saved onboarding preferences.",
tokenBudget: 1600
},
{
model: "gpt-5.5",
input: "What should we send before the implementation call?"
}
);
// Send `request` with your OpenAI client:
// await openai.responses.create(request)
The SDK places Neutron context in instructions and current user input in input.
Remote MCP
OpenAI-compatible MCP hosts can connect directly to Neutron MCP tools when your token allows it.
const { request } = await connector.createProviderContextRequest(
"openai",
{ task: "Prepare customer context." },
{
model: "gpt-5.5",
input: "Draft the next reply.",
mcp: {
serverUrl: "https://mcp.neutronai.dev/mcp",
serverLabel: "neutron_memory",
allowedTools: ["memory_agent_context", "memory_recall"],
requireApproval: "never"
}
}
);
Use read-only MCP tools for untrusted hosts. Enable write tools only for systems allowed to create, update, or delete memory.
Azure OpenAI
For Azure OpenAI, build the same request with the azure_openai adapter and send it through your Azure client or gateway.
const { request } = await connector.createProviderContextRequest(
"azure_openai",
{ task: "Summarize account memory for a support agent." },
{
model: process.env.AZURE_OPENAI_DEPLOYMENT,
input: "What should the agent know?"
}
);
Verify remote MCP network and policy support in your Azure environment before enabling MCP.