Prismor: Open-source runtime control plane for AI agents
PrismorSec released Prismor, a free open-source runtime policy layer that allows, warns, or blocks AI coding agent tool calls before execution.
Prismor sits between agents like Claude Code, Codex, or Cursor and their tool calls, enforcing allow/warn/block verdicts with 0.8 milliseconds added latency per call measured across 10,000 simulated agent sessions. It wraps npm and pip to score package installs, blocking scores of 60 or higher or any known IOC match, and ships rules covering the May 2026 mini-shai-hulud and AntV hijacked-maintainer attacks. An optional semantic guard routes uncertain prompt-injection cases to an LLM, improving catch rate by 30% on 800+ test cases without extra false positives.
- Open-source control plane intercepts Claude Code, Codex, and Cursor tool calls.
- Observe-then-enforce workflow; 0.8 ms overhead per tool call measured.
- Package manager wrapper blocks installs scoring 60+ or matching known IOCs.
- Semantic guard adds 30% prompt-injection catch rate over pattern rules.
- Presets range from 31% coverage/9% friction to 100% coverage/90% friction.
- Organizations
- PrismorSecPrismorCursor
- AI models
- Claude CodeCodex
From the original article
2 min read at helpnetsecurity.com
Prismor is a free, open-source security layer for AI coding agents. It sits between an agent such as Claude Code, Codex, or Cursor and the actions that agent wants to take, and it checks each tool call against a policy before the call runs. Every call gets one of three verdicts: allow, warn, or block.
By Anamarija Pogorelec · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.