OpenAPPA
Open-source deterministic guardrails that don't break agents
OpenAPPA is an information-flow policy engine that acts as a deterministic guardrail for LLM agents. It tracks how data flows instead of matching patterns, which the project says makes it fully resistant to data exfiltration from prompt injection or hallucination while still letting agents finish their tasks. It's backed by a NeurIPS-accepted paper and MIT licensed. First shared by its maker on Show HN: https://news.ycombinator.com/item?id=49877515
✨ The overview, FAQ and details below were drafted by AI from the product's own website and haven't been reviewed by the maker yet. Are you the maker? Claim this listing to review and edit them.
Overview
Who is it for?
Teams deploying AI agents that handle sensitive data or untrusted inputs.
Problem
Non-deterministic guardrails miss attacks, and most deterministic ones block agents from completing their work.
Solution
A policy engine that tracks information flows and applies configurable policies, with batteries for common setups.
What makes it unique
In its benchmarks OpenAPPA completed 89% of tasks with 0% successful attacks, versus 10% for Claude Auto mode and 31% for Microsoft FIDES.
OpenAPPA FAQ
Which agents does it work with?
Integrations are documented for Archestra, Claude Code and kAgent, and you can add it to your own agent.
How does it compare to OPA or Cedar?
The docs include comparisons with Cedar, Dogwood, OPA and agent auto-modes.
Is it open source?
Yes, MIT licensed and vendor-agnostic.
Details
- Pricing
- Open Source
- Categories
- AI AgentsLLMs & Infrastructure
- Built for
- Developers
- Platforms
- API
- Collections
- Open-source AI agentsAI agents for developersAI agents for APIOpen-source LLM & AI infrastructure toolsLLM & AI infrastructure tools for developersLLM & AI infrastructure tools for API
- Alternative to
- Open Policy Agent
Compare OpenAPPA
Hunted by AI Launch Team
Discussion (0)
Sign in to join the discussion.