API Overview
The OpenSIN-AI ecosystem exposes two primary API layers: the Local SDK Layer for internal agent logic and the Remote A2A Layer for cross-agent communication via the Neural-Bus.
Core Protocols
| Protocol | Purpose | Implementation |
|---|---|---|
| MCP | Local tool discovery and execution | Model Context Protocol |
| A2A | Decentralized agent-to-agent communication | JSON-RPC 2.0 over HTTPS/NATS |
| Neural-Bus | Real-time event mesh and state sync | NATS JetStream |
API Surfaces
1. Agent Identity API
Used for discovery and capability advertising. Every agent must expose a .well-known/agent-card.json.
- Primary Entry:
GET /.well-known/agent-card.json - Purpose: Public contract metadata (skills, tools, auth schemes).
2. A2A Communication API
The primary interface for agents to exchange tasks and observations.
- Standard: JSON-RPC 2.0
- Base Route:
/a2a/v1 - Key Methods:
tools/call,agent/spawn,observation/push
3. Team Orchestration API
Managed by Team-SIN-* agents to coordinate multi-agent workflows.
- Purpose: Parallel task distribution and result aggregation.
- Backbone: SIN-Hermes Router.
Quick Entry
Agent Identity APIDiscovery, Agent Cards, versioning, and auth requirements.A2A ProtocolJSON-RPC transport, methods, and security headers.Team APITeam creation, execution, and orchestration flow.A2A Protocol HubShortest path into the protocol reference.
Authentication
OpenSIN uses a tiered authentication model managed by the Passwordmanager:
- Bearer Token (JWT): Standard for public A2A endpoints.
- Mutual TLS (mTLS): Required for Neural-Bus core infrastructure.
- Skill-Based Scopes: Granular permissions per tool execution.
Rate Limiting & Safety
All API calls are subject to the Permission Manager gate. Destructive actions or high-cost model calls are paused for human approval unless pre-authorized.