⚡ New — Kimi K3 is live: bring your own Moonshot key →

Guides

Short, task-focused walkthroughs — each under two minutes. Prefer reading? Every guide links to the full docs, and agents get the same steps viallms.txt andMCP.

Get your first key

14s

Sign in with GitHub or Google, mint a key, and make your first call.

Read the docs →

Swap one line

15s

Point your OpenAI base_url at BharatRouter — same SDK, same code, from any stack.

Read the docs →

Choose how to route

12s

One knob — price, latency, uptime, throughput or auto — or set optimize_weights to dial the exact mix. Default or per call.

Read the docs →

Compare models side by side

18s

Fire one prompt at several models at once and compare answers, latency and ₹ cost — sovereign open-weight next to frontier, Sangam in the mix — then wire the winner in.

Open the playground →

Consensus routing with Sangam

16s

Send one prompt to a panel of models that answer in parallel; a synthesiser reconciles a single answer — open-weight or frontier, BYOK-ready, India-resident as an option.

Read the Sangam docs →

Route with india_only

14s

Pin requests to India-resident routes and optimize for price, latency or uptime.

Read the docs →

Save a BYOK key

12s

Bring your own Krutrim or OpenAI key — encrypted once, used on every request.

Read the docs →

Register your own endpoint (BYOE)

13s

Plug your own OpenAI-compatible deployment in as a ₹0 fallback step.

Read the docs →

Build a fallback chain & collection

14s

Chain providers for reliability, then publish it as a shareable collection.

Read the docs →

Failover that just works

13s

When a provider errors, traffic re-routes to the next healthy step — no downtime.

Read the docs →

Connect via MCP (Claude Code)

13s

Add BharatRouter as an MCP server so agents can discover and call models.

Read the docs →