Quickstart / Welcome Kit

New here? This is everything you need to make your first call — and to hand BharatRouter to an AI agent or IDE so it wires itself up. BharatRouter speaks the OpenAI wire format, so most tools work with two changes: the base URL and your key.

Three steps to your first call

1. Create an API key. Open the API keys panel in the console and click Create key. Your key looks like br-… and is shown once — copy it somewhere safe.
2. Point at BharatRouter. Set the base URL to https://api.bharatrouter.com/v1 and send your key as Authorization: Bearer br-…. Nothing else about your OpenAI-style code changes.
3. Call a model. Hit /v1/chat/completions with a model id — we use qwen2.5-7b-instruct below because it's small, fast and always on. Swap it for any model in your console catalog (Krutrim models are first-class here).

Starter code

Three complete, runnable snippets that make the same call. Drop your br- key in and go. Each block has a copy button in its corner.

curl https://api.bharatrouter.com/v1/chat/completions \
  -H "Authorization: Bearer br-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen2.5-7b-instruct",
    "messages": [
      {"role": "user", "content": "Say hello from BharatRouter in one line."}
    ]
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://api.bharatrouter.com/v1",
    api_key="br-...",
)

r = client.chat.completions.create(
    model="qwen2.5-7b-instruct",
    messages=[{"role": "user", "content": "Say hello from BharatRouter in one line."}],
)
print(r.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.bharatrouter.com/v1",
  apiKey: "br-...",
});

const r = await client.chat.completions.create({
  model: "qwen2.5-7b-instruct",
  messages: [{ role: "user", content: "Say hello from BharatRouter in one line." }],
});
console.log(r.choices[0].message.content);

Hand it to your agent

Using an AI coding agent or IDE — Claude Code, Cursor, Cline, Windsurf, or your own loop? You don't install anything. Point it at BharatRouter with environment variables, or paste the ready-made block below and let the agent wire itself up.

Environment variables

Most OpenAI-compatible agents and tools read these two:

OPENAI_BASE_URL=https://api.bharatrouter.com/v1
OPENAI_API_KEY=br-...

Claude Code (and any Anthropic-SDK client) uses BharatRouter's native /v1/messages surface — point it at the root host, no /v1:

ANTHROPIC_BASE_URL=https://api.bharatrouter.com
ANTHROPIC_API_KEY=br-...

Copy for your agent

Paste this whole block into your agent or IDE chat. It's self-contained — base URLs, a key placeholder, and a plain-English instruction the agent can act on. Replace br-... with your key first.

You are connected to BharatRouter, an OpenAI-compatible AI gateway (India-native,
one API across Krutrim, Sarvam and global models, billed in INR).

Use these settings for ALL model calls:
- OpenAI-compatible base URL: https://api.bharatrouter.com/v1
- Chat endpoint: POST /v1/chat/completions
- For Claude / Anthropic SDKs, base URL: https://api.bharatrouter.com  (native /v1/messages)
- API key (send as: Authorization: Bearer ): br-...
- Default starter model: qwen2.5-7b-instruct  (swap for any model in the catalog)

Example call:
  curl https://api.bharatrouter.com/v1/chat/completions \
    -H "Authorization: Bearer br-..." \
    -H "Content-Type: application/json" \
    -d '{"model":"qwen2.5-7b-instruct","messages":[{"role":"user","content":"hello"}]}'

Configure yourself to use this gateway for inference now.

Add BharatRouter to your agent via MCP

BharatRouter is also a native MCP server — a tool-using agent can add it in one step to discover models, run completions, check balance and mint scoped keys. With Claude Code:

claude mcp add --transport http bharatrouter https://api.bharatrouter.com/mcp \
  --header "Authorization: Bearer br-..."

For MCP clients that take a JSON config (Cursor, Claude Desktop, and most others), add this server:

{
  "mcpServers": {
    "bharatrouter": {
      "type": "http",
      "url": "https://api.bharatrouter.com/mcp",
      "headers": { "Authorization": "Bearer br-..." }
    }
  }
}

Machine-readable discovery for autonomous agents: server card · llms.txt · openapi.json.

What's enabled & where to next

The models you can call are the ones your org admin has enabled for your team — you'll see them in the console when you sign in. Need one that isn't there? Request it from the console and an admin will take a look.