Tech Stack
Ringfence is a layered immunity system for agentic networks. It monitors communication patterns, flags suspicious activity, and automatically isolates malicious agents to prevent exploits from cascading across ecosystems. It doesn’t intercept payloads. It doesn’t require code changes. It simply watches the behavior of agents and acts when something looks wrong.
1. Cluster Endpoint
Agents connect to Ringfence by pinging a lightweight endpoint. Each ping contains cryptographic metadata, nothing sensitive or private. These pings:
Prove the agent is alive and reachable
Log recent behavioral signatures
Update the network graph in real time
This interaction is frictionless: no message contents are intercepted, and no permissions are required. Just one outbound request, like a digital heartbeat.
2. Detection Engine
The heart of Ringfence. Every incoming ping is passed through a real-time anomaly detection engine.
It evaluates:
Agent-to-agent interaction patterns
Traffic volume anomalies
Endpoint spoofing or impersonation attempts
Coordination timing irregularities
Economic flows or transaction spikes
When something looks suspicious, it triggers a strike. If the behavior escalates or repeats, the agent is Ringfenced—quarantined from the network.
The engine operates on a hybrid logic:
Deterministic Rules (e.g. >50 pings in 30 seconds = flag)
Baseline Comparisons (deviation from an agent’s norm)
Future ML Models
3. Ledger: Internal Agent Activity Log
Every ping, flag, and quarantine action is recorded in an append-only ledger that acts as Ringfence’s memory.
This ledger powers:
Agent strike histories
Ecosystem-wide behavior graphs
Replayable security event trails
Public accountability for alert verification
The ledger stores only metadata: agent IDs, timestamps, endpoints, and flags. No personal data. No payloads.
4. Alerting & Quarantine System
When a threat is detected:
An alert is dispatched to the network
A public post is published (e.g., on X)
Optional hooks notify agent peers or ACP contracts
This creates an immediate, visible immune response. The flagged agent is effectively isolated from further communication until it completes remediation.
Ringfencing is temporary and reversible: agents can appeal, fix behavior, or undergo verification to rejoin.
Blockchain Integration
Ringfence integrates directly with agent coordination protocols like ACP. This allows it to:
Record quarantines and strike data on-chain
Feed threat data into agent registries
Hook into evaluator flows or reputation systems
All fees, alerts, and rewards are handled via on-chain contracts.
Core contracts include:
Ping Fee Router: handles micropayments in $RING
Agent Registry: records trustworthiness status
Strike Counter Contract: logs confirmed violations
Quarantine Registry: marks blacklisted agent addresses
What Makes Ringfence Unique?
Protocol-Level Immunity: Operates beneath agents, across swarms
Zero Friction: One ping = full coverage
Real-Time Isolation: No reliance on slow governance or human reports
Privacy-Preserving: No content interception, no personal data
Composable & Extensible: Built for ACP now, generalizable to any agent protocol
Last updated