OxDeAI signs agent actions before execution to enforce fail-closed policy boundaries
docybo · reddit · 2026-07-21
OxDeAI
OxDeAI is an open-source protocol for deterministic pre-execution authorization in AI agents.
Core idea
Instead of checking policy inside the agent loop, the agent proposes an intent and OxDeAI deterministically evaluates (intent, state, policy). If it allows the action, it emits a signed AuthorizationV1 artifact that a guard verifies before any side effect can happen.
What exists today
- signed decision artifacts and a non-bypassable guard
- adapters for LangGraph, CrewAI, AutoGen, OpenAI Agents SDK, and OpenClaw
- single-hop scoped delegation
- cross-language conformance vectors for TS, Go, and Python
- hash-chained audit envelopes for offline verification
Honest caveats
- cross-language coverage is not complete for every verdict yet
- benchmark numbers are only indicative
- there is no third-party security review yet
- the protocol is still evolving
The author explicitly says this is not a prompt guardrail or monitoring tool; it enforces the execution boundary.
Related event: OxDeAI Open-Source Protocol Sets Execution Boundaries for AI Agents(2 posts)→
More from coding & agent
- 'My intelligence stack': developer splits AI coding work across Opus 5.5, Fable and Astra with human review gates — sh_reya · 2026-10-27
- Baseten webinar: evaluate open models for coding agents by benchmarking on your own merged PRs — baseten · 2026-10-27
- Thorsten Ball on working with agents: run more evals, automate everything, use the best models — omarsar0 · 2026-10-27
- Where to draw the approval line for scheduled agent jobs — Reddit discussion — MemStrata · 2026-10-27
- 4 MCP servers worth a look: full-text YouTube search, Chrome session reuse, Framer editing — Prior_Engineering486 · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27