PrismorSecSecurity & privacy
prismor
A self-hosted runtime security control plane for AI coding agents, providing real-time monitoring, blocking, and human-in-the-loop approval for tool calls.
- DevOps
- Security
- Backend
- Automation
- Monitoring / Observability
- CLI
- Library/framework
- macOS
- Linux
- Windows
- Self-hostable
- Runs locally
- Popularity
- 325 Stars
- GitHub stars
- Recent activity
- 8/20/2026
- Updated in the last 30 days
- License
- APACHE-2.0
- Permissive
Why it matters
We look beyond stars: what problem it solves, whether it creates real utility, and what makes its approach worth noticing.
Problem
AI coding agents possess broad file access, shell execution, and API capabilities, making them vulnerable to prompt injection, secret exfiltration, and destructive actions without adequate runtime control.
Practical value
Provides a self-hosted runtime control plane via Python to observe, human-in-the-loop approve, or block tool calls across major AI coding agents like Claude Code and Codex.
Innovation / differentiation
Introduces an agent-layer-specific security policy engine, MCP gateway, and skill scanner to intercept and review tool calls and responses before execution, filling a gap left by OS-level tools.
Leverage potential
Integrates with a wide range of AI coding agents and MCP architectures like Claude Code, Cursor, GitHub Copilot, and Aider, providing a unified control layer for development environments.
Why now
As AI coding agents see widespread adoption in daily development, securing codebases, credentials, and tool executions has become an urgent real-world priority for teams.
Community activity
In the last 90 days there were 62 new issues and 163 pull requests; the bounded issue/PR samples include 4 issue authors and 3 PR contributors, with at least 50 releases.
Maintainer responsiveness
The 62-issue window sample had a 58% close rate, and the 99-pull-request sample had a 90% merge rate. Maintainer-response observations covered 98% of that issue sample, with a 0% response rate and median first response of not enough data.
Key highlights
- Provides a runtime policy engine with a local dashboard to intercept or manually approve AI agent tool calls
- Supports supply chain security checks for package installations, IOC matching, and risk scoring
- Enplements network isolation with policy-driven egress control, raw IP detection, and tunnel blocking
Quick start
How it is installed, how hard it is, and where to start.
Where it runs
Runs locally
Difficulty
Medium — some setup needed
- 01Install the Prismor core package using pip.
- 02Run the setup command to configure security policies.
- 03Launch or wrap your AI agent with Prismor to begin monitoring.
Best for
- Teams that want data on their own servers
- Developers who want to try it on their machine
More about it
Prismor is a runtime security control plane designed for AI coding agents. It establishes a protective layer between agents and system resources, monitoring and managing shell command execution, file I/O, and API calls to mitigate risks such as prompt injection, privilege escalation, and supply chain attacks.
Supporting a wide range of AI agent frameworks, Prismor allows users to implement 'observe' or 'enforce' modes via custom policies. Key features include an MCP gateway, semantic guardrails, honeytoken credential planting, and Ed25519-signed audit trails to ensure agent behavior is traceable and compliant.
Sources
Each field shows its status and source — expand to review.
14 · Expand
Sources
Each field shows its status and source — expand to review.
capability tags
Verifiedsecurity, automation, monitoring
Source: admin_cms · cms editor · 8/21/2026
editor note
Verified{"en":"Prismor intercepts tool calls at the agent layer rather than relying on OS-level monitoring, effectively preventing malicious AI actions. It is well-suited for development teams integrating AI coding assistants into production environments with strict permission requirements.","zh":"Prismor 通过在代理层拦截工具调用而非仅依赖操作系统监控，能有效防止 AI 代理执行恶意操作。它特别适合需要将 AI 编码助手接入生产环境并进行严格权限管控的开发团队。"}
Source: admin_cms · cms editor · 8/21/2026
how to use
Verified{"steps":[{"en":"Install the Prismor core package using pip.","zh":"使用 pip 安装 Prismor 核心包。"},{"en":"Run the setup command to configure security policies.","zh":"运行初始化命令以配置安全策略。"},{"en":"Launch or wrap your AI agent with Prismor to begin monitoring.","zh":"通过 Prismor 启动或挂载 AI 代理以开始监控。"}],"installAt":"local","difficulty":"medium"}
Source: admin_cms · cms editor · 8/21/2026
intro
Verified{"en":"Prismor is a runtime security control plane designed for AI coding agents. It establishes a protective layer between agents and system resources, monitoring and managing shell command execution, file I/O, and API calls to mitigate risks such as prompt injection, privilege escalation, and supply chain attacks.\n\nSupporting a wide range of AI agent frameworks, Prismor allows users to implement 'observe' or 'enforce' modes via custom policies. Key features include an MCP gateway, semantic guardrails, honeytoken credential planting, and Ed25519-signed audit trails to ensure agent behavior is traceable and compliant.","zh":"Prismor 是一个专为 AI 编码代理设计的运行时安全控制平面。它通过在代理与系统资源之间建立防护层，监控并管理代理的 shell 命令执行、文件读写及 API 调用，从而降低提示词注入、权限提升及供应链攻击风险。\n\n该项目支持多种主流 AI 代理框架，允许用户通过自定义策略实现「观察模式」或「强制执行模式」。其核心功能包括 MCP 网关、语义防护、蜜罐凭证植入以及基于 Ed25519 签名的审计追踪，确保 AI 代理的行为可追溯且符合安全规范。"}
Source: admin_cms · cms editor · 8/21/2026
Latest release
Verifiedv1.43.0
Source: GitHub API · latest_release=v1.43.0 · 8/21/2026
License
VerifiedApache-2.0
Source: GitHub API · license.spdx_id=Apache-2.0 · 8/21/2026
needs api key
VerifiedNo
Source: admin_cms · cms editor · 8/21/2026
One-liner
Verified{"en":"A self-hosted runtime security control plane for AI coding agents, providing real-time monitoring, blocking, and human-in-the-loop approval for tool calls.","zh":"用于 AI 编码代理的自托管运行时安全控制平面，支持对代理工具调用进行实时监控、拦截及人工审核。"}
Source: admin_cms · cms editor · 8/21/2026
Platforms
Verifiedmacos, linux, windows
Source: admin_cms · cms editor · 8/21/2026
Category hint
Inferred from materialsai-apps
Source: Project README · hint=ai-apps · 8/16/2026
product forms
Verifiedcli, library_framework
Source: admin_cms · cms editor · 8/21/2026
role tags
Verifieddevops, security, backend
Source: admin_cms · cms editor · 8/21/2026
supports local
VerifiedYes
Source: admin_cms · cms editor · 8/21/2026
supports self host
VerifiedYes
Source: admin_cms · cms editor · 8/21/2026
Related projects
Other verified projects matched by category, capabilities, and intended roles.
agent-governance-toolkit
A governance toolkit for AI agents that adds policy checks, identity, audit trails, sandboxing, and SRE controls across multiple agent frameworks.
fleet
A device-management platform for IT and security teams covering MDM, software deployment, patching, inventory queries, and compliance across desktop and mobile operating systems.
openbao
An open-source secrets management system for storing, distributing, rotating, and controlling access to secrets, certificates, and keys.
osv.dev
Open source vulnerability database and triage service providing a unified query API and web interface
CVE Lite CLI
A local-first JavaScript/TypeScript lockfile vulnerability scanner focused on actionable remediation, with CI, SARIF, HTML reports, and offline advisory data.
VirusDetector
A Manifest V3 Chrome/Edge extension that detects Silver Fox Trojan phishing and spoofed websites using multi-rule scoring and download blocking.