Run your first agent

Connect a model provider

A provider holds the model key your agents use. Blazing Agents stores the key for you, so your program sends it only once. Create quickstart.ts or quickstart.py with this code:

import { BlazingAgents } from "@blazingagents/sdk";

const client = new BlazingAgents({
  apiKey: process.env.BLAZING_AGENTS_API_KEY!,
});

const providerName = "Quickstart OpenRouter";
const { providers } = await client.providers.list();
const provider =
  providers.find(({ name }) => name === providerName) ??
  (await client.providers.create({
    name: providerName,
    providerType: "openrouter",
    baseUrl: null,
    apiKey: process.env.OPENROUTER_API_KEY!,
  }));
console.log(`Provider: ${provider.id}`);

Run it with node quickstart.ts or python quickstart.py. You see Provider: prv_..., and the same ID on every later run.

Create an agent

An agent pairs a provider with a model. Add this to the end of the file:

const agentName = "Quickstart agent";
const { agents } = await client.agents.list();
const agent =
  agents.find(({ name }) => name === agentName) ??
  (await client.agents.create({
    name: agentName,
    providerId: provider.id,
    model: "openai/gpt-6-luna",
  }));
console.log(`Agent: ${agent.id}`);

Run the file again. You see the same Provider: prv_... line, then Agent: ag_....

Blazing Agents checks the model against your provider when you create the agent. If you get model_not_found, pick another ID from client.providers.listModels({ providerId }) in TypeScript or client.providers.list_models(provider_id=...) in Python. The first step stores your OpenRouter key without checking it, so a wrong or invalid key shows up here as model_validation_unavailable. Fix OPENROUTER_API_KEY, delete the "Quickstart OpenRouter" provider with client.providers.delete({ providerId }) in TypeScript, client.providers.delete(provider_id) in Python, or the Delete action on the dashboard's providers page, then run the file again to recreate it with the new key.

Stream the first answer

Send the agent a message and print the answer as it arrives. Add this to the end of the file:

const result = await client.chat({
  agentId: agent.id,
  message: {
    id: crypto.randomUUID(),
    role: "user",
    parts: [{ type: "text", text: "Say hello in one short sentence." }],
  },
});
console.log(`Session: ${await result.sessionId}`);

for await (const chunk of result.toStream()) {
  process.stdout.write(chunk);
}
// The stream is yours now: relay it to your app, parse it, or store it.

Run the file once more. After the IDs, the raw stream prints as it arrives:

Provider: prv_...
Agent: ag_...
Session: ss_...
data: {"type":"start",...}

data: {"type":"text-delta","id":"...","delta":"Hello"}

data: {"type":"text-delta","id":"...","delta":"!"}

...

data: [DONE]

These lines are AI SDK UI message events. In a web app your backend returns result.toResponse(), and the AI SDK useChat hook renders them as chat messages. Each run starts a new session, so you see a new ss_... ID every time.