pic-standard
Active·★ 68·Apache-2.0·Updated 2026-10-21
PIC Standard is a lightweight, local-first protocol that forces AI agents to prove every important action before it happens, verifying intent, impact, provenance, and evidence. It acts as an action-bound verification contract that fails closed if any aspect of a high-impact tool call is found to be wrong or unjustified.
#AI Agent Security#Action Gating#Provenance Verification#Evidence Verification#Local-first Protocol#Agent Safety#Tool Call Guarding#Prompt Injection Prevention
01
Features
01Pre-execution verification of AI agent actions
02Local-first operation with zero cloud dependency
03Prevents prompt injection and blind tool calls
04Integrates with major AI agent frameworks (LangGraph, MCP, OpenClaw, Cordum)
05Supports deterministic evidence (hash, signature) verification
02
Compatibility
Python
Runtime
Verified via docs
LangGraph
Integration
Verified via docs
MCP
Integration
Verified via docs
OpenClaw
Integration
Verified via docs
Cordum
Integration
Verified via docs
HTTP Bridge
API Interface
Verified via docs
03
Quick start
1
$ pip install pic-standard
04
Use cases
↳Guarding high-impact actions within AI agent frameworks
↳Securing internal tool gateways from unauthorized agent actions
↳Preventing AI hallucinations from leading to real-world damage (e.g., unauthorized wire transfers)
↳Ensuring verifiable provenance and evidence for critical agent decisions
↳Implementing fail-closed action policies for enhanced agent safety
05
Alternatives
ADR★ 2.1k
ADR (Agentic Detection and Response): open-source telemetry sensor and dual-agent detection framework for securing MCP-based AI agents.
AgentHound★ 452
Offensive security framework for AI agent infrastructure - recon, credential looting, model exfiltration, poisoning, and attack-path analysis across MCP, A2A, gateways, and AI services. BloodHound for the agentic stack.
helm-ai-kernel★ 60
Fail-closed execution firewall for AI agents: quarantine MCP tools, proxy OpenAI-compatible requests, emit signed receipts, and verify EvidencePacks offline.
ephemora-cell★ 48
Ephemora Cell — The execution layer for untrusted AI-generated code. Fast, capability-based WASM execution with explicit CPU, memory, time, I/O, and filesystem limits.
agentshield★ 21
🛡️ The security firewall for AI agent tools & MCP servers (Claude, GPT-5.6, Gemini 3.7, Antigravity, Cursor, Codex, Hermes). Catch command injection, secret theft & toxic flows in <50ms. 100% offline Rust SAST + 1-click auto-fix + SARIF.
phantom-secrets★ 18
Stop AI coding agents from leaking your API keys. Local proxy + MCP that swaps real secrets for phm_ tokens — works with Claude Code, Cursor, Windsurf, and Codex.
Comments
Log in to leave a comment
No comments yet. Be the first!