About

The trust layer the
agentic web needs

Aroha Labs was founded on a simple observation: AI agents are becoming capable enough to take real actions in the world — book flights, move money, submit pull requests — but the infrastructure to keep humans in control does not exist yet.

We are building that infrastructure. The Aroha Protocol gives every agent a cryptographically-signed spending mandate, a tamper-evident audit log, and a DID-anchored identity — so that when an orchestrator delegates to a sub-agent, the human who started the chain stays in control of what that sub-agent is allowed to do.

Our mission

Make AI delegation safe, auditable, and human-controlled — for every developer, in every language, on every platform.

What we stand for

Human control, always

Every delegation chain must trace back to a human approval. We build protocols that make this the default, not the exception.

Open by default

The Aroha Protocol is MIT-licensed. No lock-in, no proprietary extensions you can't see. If we build it, you can read it.

Security through legibility

Agents that can't explain what they did can't be trusted. Every Aroha response carries a verifiable audit trail — no black boxes.

Interoperability over dominance

We ship bridges to LangChain, CrewAI, MCP, and Google A2A because a fragmented agentic web helps nobody.

Open source

The Aroha Protocol, all four SDKs, the registry server, and this website are open-source under the MIT License. We believe the trust layer for the agentic web should be inspectable by anyone who depends on it.

SDK downloads, conformance test results, and community contributions are public.

Why the protocol can't be forked away

The Aroha Protocol is MIT-licensed — anyone can fork it. The moat isn't the code, it's the network.

Reputation data accretes over every call

The Bayesian reputation graph learns from every transaction — response time, mandate compliance, satisfaction signals. A fork starts with zero history. Aroha's Hub accumulates it permanently.

Two-sided network lock-in

Vendors integrated on the Hub don't re-integrate on a fork. Agent builders don't switch registries. Both sides wait for the other — the first-mover advantage is compounding.

Protocol governance moves to a foundation

The spec is being contributed to an open foundation (modeled on the Linux Foundation's MCP stewardship). That institutional legitimacy can't be forked.

Start building

The fastest way to understand Aroha is to run a live agent in under five minutes.