Frequently Asked Questions¶

What is the primary objective of NANDA?¶

NANDA is an open-source initiative and research organization advancing decentralized, autonomous multi-agent architectures for public infrastructure, crowd assistance, and civic intelligence.

How does NANDA differ from general LLM workflow frameworks?¶

Conventional frameworks focus primarily on sequential prompt chaining and API orchestration on cloud servers. NANDA is built specifically for edge-to-cloud mesh communication, formal verification, zero-knowledge audit trails, and real-world physical crowd deployments where fault tolerance and local fallback execution are required.

Under what license is NANDA distributed?¶

All core repositories, SDKs, benchmarks, and reference implementations are released under permissive open-source licenses (Apache 2.0 and MIT).

Can NANDA agents execute without internet access?¶

Yes. Field deployments utilize quantized Small Language Models (SLMs) and localized state caches running on edge hardware gateways. Agents can communicate over local Bluetooth or Wi-Fi mesh networks and sync with the broader network once connectivity is restored.

How can a municipal administration or institution partner with NANDA?¶

Institutional inquiries regarding pilot deployments, research partnerships, and testing testbeds can be submitted via email to partnerships@nanda.ai.

How are benchmarks published on the NANDA Index verified?¶

Submissions must include deterministic evaluation traces generated by the nanda-benchmark CLI. Traces are verified against synthetic adversarial inputs to compute standardized safety and latency scores.