Pipelock

CLI Verified Active

Open-source AI agent firewall for MCP security: agent egress control, DLP, SSRF, and prompt injection defense.

Purpose

To secure AI agent workflows by acting as an inline firewall, protecting against data exfiltration, prompt injection, and other threats, while providing verifiable proof of agent actions.

Features

AI agent firewall with mediator-signed receipts
Process containment via OS-native primitives
MCP security, egress control, DLP, and prompt injection defense
11-layer URL scanner with advanced evasion detection
Audit logging and verifiable evidence emission

Use Cases

Securing AI agents with sensitive credentials in their environment.
Preventing prompt injection attacks in LLM interactions.
Auditing and verifying the actions taken by AI agents.
Integrating AI agent security into CI/CD pipelines.

Non-Goals

Replacing fundamental OS security controls.
Providing agent-side vulnerability scanning.
Acting as a general-purpose network proxy without security focus.

Quality Score

Verified

100 /100

Analyzed 14 days ago

Trust Signals

Last commit14 days ago

GitHub owner luckyPipewrench

Stars584

LicenseApache-2.0

Websitepipelab.org

Status

View Source

Similar Extensions

Speclock Ros2

100

AI Constraint Engine — enforces CLAUDE.md, .cursorrules, AGENTS.md rules as laws. 51 MCP tools, 991 tests. Official MCP Registry. npx speclock protect

CLI

sgroy10

Demo Metrics

Open-source AI agent firewall for MCP security: agent egress control, DLP, SSRF, and prompt injection defense.

CLI

luckyPipewrench

@codeharbor/agent Playbook

100

Local skill manager and installer for agent-playbook across Claude Code, Codex, and Gemini.

CLI

charon-fan