SASY releases open-source Datalog policy engine for AI agent tool calls
Start here
SASY released an Apache-2.0 SDK and policy engine that checks each AI agent tool call against rules written in Datalog before the call executes. It denies actions whose history traces back to both untrusted and sensitive documents. Policies are compiled to native code, so decisions are deterministic rather than made by an LLM, and an analyzer flags conflicting or redundant rules. The SDK hooks automatically into LangChain, Google ADK and Langroid; custom agents call sasy.record and sasy.check_tool_call, and setup requires a Python project with uv and Docker installed. A separate SASY Guard applies the same checks to Claude Code, Codex and Pi, with Claude Code needing version 2.1.289 or later.
Generated from reports · updated 2 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AISASY launches open-source Datalog policy engine that checks every AI agent tool call
SASY released an Apache-2.0 SDK and engine that checks each AI agent tool call against policies written in Datalog before the call runs, denying actions whose history traces back to both untrusted and sensitive documents. The SDK hooks into LangChain, Google ADK and Langroid automatically, and a separate SASY Guard brings the same checks to Claude Code, Codex and Pi. Policies are compiled to native code, decisions are deterministic rather than LLM-judged, and an analyzer flags conflicting or redundant rules.
Heat for this event
Not enough continuous observations yet to draw a trend.