SHACKLE Protocol SP/1.0: Open-Source Runtime Circuit Breaker for AI Agents Launches

Summary

Sovereign Logic has published the SHACKLE Protocol Specification (SP/1.0), an open-source runtime circuit breaker designed to prevent autonomous AI agents from entering infinite loops and wasting resources through unchecked tool execution. The specification addresses a critical gap in current AI agent frameworks, where agents can enter retry loops or spawn zombie processes that consume thousands of dollars in API costs before hitting framework-level guardrails.

SHACKLE operates as either a sidecar daemon with gRPC/Unix socket transport or as an in-process library, providing deterministic verdicts based on a pure decision function backed by 9 mathematical invariants. Every decision is Ed25519-signed and logged to an immutable audit chain, enabling cryptographic verification of authorization decisions. The system operates on a fail-closed principle: network failures or daemon timeouts automatically deny execution unless explicitly authorized.

The reference implementation is dual-licensed under AGPLv3 (open source) and commercial terms, with the specification published under Creative Commons Attribution 4.0. This represents the industry's first open-source runtime circuit breaker for AI agents with cryptographic audit chain-of-custody, and follows an emerging consensus among AI safety teams that generation authority (what the model chooses) must be separate from release authority (what the system allows to execute).

• Available as sidecar daemon (production) or in-process library (single-agent), with Redis-backed distributed state engine for scaling
• Dual-licensed under AGPLv3 and commercial terms; represents industry shift toward treating agent authorization as a separate mediation layer from model generation