PrismorSec/prismor — Prismor (formerly Immunity Agent) - runtime security for Claude Code, Cursor, Windsurf & other AI co…

Description and use

**Brief Summary**
Prismor (formerly Immunity Agent) is an open‑source runtime security layer for AI coding assistants such as Claude Code, Cursor, and Windsurf. It intercepts tool‑use calls with PreToolUse hooks to block dangerous commands, prevent secret leaks, stop prompt‑injection attacks, and gate risky package installations, all via a standard Model Context Protocol (MCP) interface.

**Value**
Prismor gives developers a turnkey way to harden AI‑driven code‑generation pipelines without rewriting the agents themselves. By exposing a uniform MCP/SDK/CLI, it lets you plug security checks into any supported AI coder, enforce organization‑wide policies (e.g., no external network calls, no secret exposure), and safely expose real tools and databases to the model.

**Practical Adoption Path**

1. **Prototype** – Clone the repo, install the Python package, and run the provided MCP server locally.
2. **Integrate** – Register the server as a “tool” in your AI agent’s configuration (most agents already support MCP hooks).
3. **Policy Definition** – Write simple JSON/YAML rules that specify which commands, packages, or environment variables are blocked or require approval.
4. **Test** – Use the built‑in CLI to simulate tool calls and verify that unsafe actions are intercepted.
5. **Roll Out** – Deploy the MCP server as a container or managed service, point production agents at it, and optionally enable a human‑in‑the‑loop approval UI for high‑risk actions.

**Production Readiness**

• **Activity & Adoption**: 214 ★, recent commits (last updated 2026‑06‑29), and growing ecosystem usage indicate a healthy community.
• **Maturity**: Core functionality (PreToolUse hooks, secret‑leak detection, package‑gate) is stable; the Python implementation is well‑documented with SDK and CLI entry points.
• **Risk Profile**: No critical metadata issues have been identified, though a final review of the license and maintainer responsiveness is advisable.

Overall, Prismor is ready for a serious pilot in production environments, especially where AI‑assisted coding must meet strict security and compliance requirements.

Problem

PrismorSec/prismor helps connect AI assistants to real tools and data through a standard protocol.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• connect AI agents to tools
• ship Model Context Protocol servers
• standardize integrations

Evidence

Prismor (formerly Immunity Agent) - runtime security for Claude Code, Cursor, Windsurf & other AI coding agents. PreToolUse hooks that block dangerous commands, prevent secret leaks, stop prompt injection, and gate risky package installs.

Source