| title | Protocol Bridges |
|---|---|
| description | 12 protocol bridges connecting SINT to robotics, industrial, and AI ecosystems. |
| sidebarTitle | Bridges |
SINT Protocol includes 12 bridge adapters that translate between the SINT capability token system and external protocols. Each bridge enforces the same security guarantees — capability verification, constraint checking, and evidence logging — regardless of the downstream protocol.
Every bridge follows the same pattern:
Agent Request → Capability Token → Policy Gateway → Bridge Adapter → External Protocol
↕
Evidence Ledger
Bridges are not passthrough proxies. They intercept every command, validate the agent's capability token, enforce physical constraints, and log the decision before forwarding to the target system.
**Package:** `@sint/bridge-ros2`
Bridges SINT capability tokens to ROS2 (Robot Operating System 2). Wraps `cmd_vel`, joint commands, and sensor subscriptions with policy enforcement.
- Control loop latency: P50 < 1ms, P95 < 2ms (benchmarked)
- Supports `geometry_msgs/Twist`, `sensor_msgs/JointState`
- Topic-level capability scoping
Connects SINT to drone and UAV systems via the MAVLink protocol. Enforces geofence constraints, altitude limits, and velocity caps.
- MAVLink v2 message framing
- GPS-based geofence enforcement
- Flight mode restriction via capability tokens
Integration with Open-RMF (Robotics Middleware Framework) for multi-robot fleet management in facilities like hospitals, warehouses, and airports.
- Fleet dispatch governance
- Zone-based access control
- Task priority enforcement
Hardware Abstraction Layer for direct actuator control. Provides a unified interface across motor controllers, servos, and sensor arrays.
- Vendor-agnostic motor API
- Real-time constraint enforcement
- Sensor fusion support
Wraps gRPC service calls with capability token authentication. Supports unary, server-streaming, and bidirectional streaming RPCs.
- 176+ lines of protobuf-driven integration
- Interceptor-based token injection
- Streaming policy enforcement
Sparkplug B protocol adapter for SCADA and industrial IoT. Bridges SINT policy enforcement to MQTT-based control systems.
- Sparkplug B topic namespace compliance
- Device/node birth/death tracking
- Metric-level access control
OPC Unified Architecture adapter for industrial automation. Connects SINT to PLCs, SCADA systems, and manufacturing execution systems.
- Node-level capability scoping
- Subscription-based monitoring with policy gating
- Historical data access governance
Generic IoT device bridge for constrained environments. Supports lightweight protocols and device shadow patterns.
- Device twin management
- Command-and-control with capability tokens
- Telemetry ingestion with policy filtering
Model Context Protocol adapter. Allows MCP-compliant AI agents to interact with SINT-governed resources. Tool calls are intercepted and policy-checked before execution.
- MCP tool discovery with capability-scoped filtering
- Context preservation across tool calls
- Automatic evidence logging per tool invocation
Agent-to-Agent protocol bridge (Google A2A). Enables secure inter-agent communication with capability token delegation.
- Agent card registration and discovery
- Message routing with capability verification
- Task delegation across agent boundaries
Metered resource management. Tracks agent budgets, computes action costs, and enforces economic constraints.
- Per-agent balance tracking
- Cost quoting before execution
- Budget enforcement with overdraft prevention
- Event stream for billing integration
Multi-agent swarm coordination. Manages consensus, task allocation, and collective decision-making for agent groups.
- Swarm membership governance
- Consensus protocol with capability-weighted voting
- Task allocation with constraint propagation
In addition to protocol bridges, SINT includes three cognitive engine packages:
| Engine | Package | Purpose |
|---|---|---|
| System 1 | @sint/engine-system1 |
Fast, reactive subsystem — handles real-time sensor responses and reflexive safety actions (sub-millisecond) |
| System 2 | @sint/engine-system2 |
Deliberative planning subsystem — handles complex multi-step reasoning and plan generation |
| Capsule Sandbox | @sint/engine-capsule-sandbox |
Isolated execution environment for untrusted agent code with syscall-level sandboxing |
All bridges implement the same interface pattern:
import { PolicyGateway } from "@sint/gate-policy-gateway";
import { LedgerWriter } from "@sint/gate-evidence-ledger";
export function createBridge(
gateway: PolicyGateway,
ledger: LedgerWriter,
config: BridgeConfig
) {
return {
async handleCommand(cmd: ExternalCommand) {
// 1. Map external command to SINT intercept request
const request = mapToIntercept(cmd);
// 2. Policy check (token validation, constraints, tier)
const decision = gateway.intercept(request);
// 3. Log to evidence ledger
ledger.append(decision);
// 4. If approved, forward to external system
if (decision.outcome === "approve") {
return forwardToProtocol(cmd, decision);
}
return { denied: true, reason: decision.reason };
},
};
}The bridge system is exposed through the gateway server's unified API:
| Bridge | Endpoint | Method |
|---|---|---|
| Core intercept | /v1/intercept |
POST |
| Batch intercept | /v1/intercept/batch |
POST |
| A2A messaging | /v1/a2a |
POST |
| A2A agent registry | /v1/a2a/agents |
GET, POST |
| Economy balance | /v1/economy/balance/:agentId |
GET |
| Economy budget | /v1/economy/budget/:agentId |
GET |
| Economy quote | /v1/economy/quote |
POST |
| Economy routing | /v1/economy/route |
POST |
| Risk stream | /v1/risk/stream |
GET (SSE) |
All bridge endpoints inherit the gateway's authentication, rate limiting, and evidence logging.