Your AI Agent Should Speak
All 22 of Your Languages
An OpenClaw fork. Powered entirely by Indian LLMs.
OpenClaw has 60K+ stars and proved open-source AI agents work. But it runs on Claude, GPT, and DeepSeek. India now has production-grade LLMs that cover 22 languages. Someone needs to fork OpenClaw and build it on top of those. That's what IndusClaw is for.
Why This Needs to Exist
Sarvam-105B has 128K context across 22 languages. Krutrim-2 does fast inference at 12B parameters. Hanooman generates video. AI4Bharat's IndicTrans2 translates across every scheduled language. The models are here. What's missing is an AI agent platform that treats them as first-class citizens.
An AI agent you can talk to in Hindi, Tamil, or any of 22 scheduled languages — natively, not through a translation layer
Indian LLMs as the default model — Sarvam, Krutrim, or Hanooman out of the box, not a fallback buried in a config file
Voice-first interaction on WhatsApp — because that's how hundreds of millions of Indians actually communicate
Self-hostable on Indian infrastructure — your data never leaves the country unless you want it to
Model-agnostic under the hood — start with Indian LLMs, plug in others when you need them
Fully open source, MIT licensed — fork it, extend it, make it yours
Model
Indian LLMs We Want to Power This
These aren't research toys. They're production models from IITs, government labs, and well-funded startups. The foundation is here — the agent platform on top is what we need to build.
| Model | Organization | Parameters | Languages | Notes |
|---|---|---|---|---|
| Sarvam-105B | Sarvam AI | 105B MoE | 22 Indian + EN | Govt-backed, open-source, 128K ctx |
| Sarvam-30B | Sarvam AI | 30B MoE | 10+ Indian | Lighter, 32K ctx, edge-friendly |
| Krutrim-2 | Ola AI | 12B dense | 22 Indian | Fast inference, 128K ctx |
| Hanooman | IIT Bombay + Jio | 1.5B–40B | 11 Indian | Multimodal: text, speech, video |
| Airavata | AI4Bharat | Various | 22 Indian | NLP tooling, IndicTrans2 |
| BharatGen | CDAC + IndiaAI | Various | 22 Indian | Govt supercomputing initiative |
Help Build This
This is day zero. No fork exists yet. We need developers, translators, and Indian LLM builders who want to make this real. Leave your email — we'll reach out when it's time to ship.
You're in. We'll reach out when it's time.