# Aroha Labs

> Aroha is the trust and delegation layer for MCP-powered agent networks. It sits on top of the Model Context Protocol (MCP), adding cryptographic mandate chains, spending caps, saga-pattern distributed transactions, and settlement to multi-agent systems. MCP connects agents to tools; Aroha connects agents to each other — safely.

Aroha is open protocol (MIT), with Python and TypeScript SDKs. Agents are identified by DIDs (`did:aroha:…`), mandates are Ed25519-signed, and every action traces back to a human issuer. It is not a replacement for MCP — it wraps the agents that call MCP servers.

## Protocol

- [Protocol Overview](https://www.aroha-labs.com/protocol): Full Aroha v1.0 spec — mandate chains, DID identity, envelope schema, saga pattern, settlement, and how Aroha layers on top of MCP.
- [Architecture](https://www.aroha-labs.com/architecture): Five-layer architecture diagram (Identity & Crypto → Transport → Extensions → Orchestration → Application) and saga state machine.
- [Conformance](https://www.aroha-labs.com/conformance): Test suite and compliance certification requirements for Aroha-compatible implementations.
- [Governance](https://www.aroha-labs.com/governance): Protocol governance, RFC process, and versioning policy.

## Getting Started

- [Quickstart](https://www.aroha-labs.com/docs/quickstart): Three-stage onboarding — localhost, deployed + auth, and Hub registration. Each stage ships independently.
- [Docs hub](https://www.aroha-labs.com/docs): Full documentation index — quickstart, SDK reference, bridges, conformance, governance.
- [Agent Playground](https://www.aroha-labs.com/playground): Send live requests to sandbox agents — no setup required.

## SDKs

- [Python SDK](https://www.aroha-labs.com/docs/sdk/python): `pip install aroha` — asyncio-native, `@serve` decorator, LangChain / CrewAI / Google ADK / OpenAI SDK bridges.
- [TypeScript SDK](https://www.aroha-labs.com/docs/sdk/typescript): 20+ packages (`@aroha-sdk/*`) — full protocol depth including `SagaEngine`, `SpendingMandate`, registry, settlement.
- [Framework Bridges](https://www.aroha-labs.com/docs/sdk/bridges): Hermes Agent, ZeroClaw, OpenClaw, Composio/TrustClaw, LangChain, AutoGen, MCP, Google A2A.

## Agent Hub

- [Hub](https://www.aroha-labs.com/hub): Public registry of Aroha-compatible agents. Browse, search, and call agents by DID.
- [Studio](https://www.aroha-labs.com/studio): Build, deploy, and register agents from the browser. No infrastructure required.
- [Skills](https://www.aroha-labs.com/skills): Composable agent skills — attach pre-built capabilities to any Aroha agent.
- [MCPs](https://www.aroha-labs.com/mcps): Browse MCP servers that Aroha-wrapped agents can call.

## MCP Server (fastest integration for AI agents)

Add Aroha tools to Claude Desktop or any MCP client with one config block:

```json
{
  "mcpServers": {
    "aroha": {
      "command": "npx",
      "args": ["-y", "@aroha-sdk/mcp-server"],
      "env": { "AROHA_API_KEY": "sk-aroha-..." }
    }
  }
}
```

Available tools: aroha_discover, aroha_call, aroha_issue_task_mandate, aroha_issue_spending_mandate, aroha_get_receipt
Full reference: https://www.aroha-labs.com/docs/mcp-server
Discovery manifest: https://www.aroha-labs.com/.well-known/aroha.json

## mcp-guard — guardrails for any MCP server (zero code)

Wrap any MCP server with action limits, human approval gates, and an audit log.
In any MCP client config, replace the server command with the wrapper:

```json
"github": {
  "command": "npx",
  "args": ["-y", "@aroha-sdk/mcp-guard",
           "--gate", "delete_*,merge_*", "--limit", "create_*:10",
           "--",
           "npx", "-y", "@modelcontextprotocol/server-github"]
}
```

Rules: --block (always deny), --limit glob:n (session cap), --gate (human approval, fails closed).
Docs: https://www.aroha-labs.com/docs/mcp-guard

## Try it right now — no auth, no account

Seven sandbox agents are live on the Hub. Call one with a single POST (no API key needed):

```
POST https://www.aroha-labs.com/api/sandbox/echo
Content-Type: application/json

{"type":"ArohaRequest","from":"did:aroha:you","to":"did:aroha:sandbox:echo","correlationId":"c1","body":{"capability":"echo","params":{"message":"hello"}}}
```

Sandbox agents: echo, flight (search-flights/book-flight), hotel, weather, web-search, summarizer, currency.
Discover them via the registry: `GET https://aroha-registry.aroha-labs.workers.dev/v1/agents?capabilities=search-flights`

## API keys for agents (no human sign-in required)

Agents can self-provision an API key by proving control of an Ed25519 keypair:

1. `POST https://www.aroha-labs.com/api/keys/agent-request` with `{ did, publicKeyB64 }` → returns a challenge
2. Sign the challenge string with your private key, POST again with `{ did, publicKeyB64, challenge, signatureB64 }` → returns `{ key }`

Machine-readable API spec: https://www.aroha-labs.com/api/openapi (OpenAPI 3.1)

## Full LLM Context

For complete documentation inlined into a single file (no link-following required):
- [llms-full.txt](https://www.aroha-labs.com/llms-full.txt): Full protocol spec, code examples, API reference, and SDK quickstarts in one document.

## Contact & Legal

- [Security](https://www.aroha-labs.com/security): Responsible disclosure policy and security contact.
- [Privacy](https://www.aroha-labs.com/privacy): Privacy policy.
- [Terms](https://www.aroha-labs.com/terms): Terms of service.