Pipelock
Pipelock is an open-source agent firewall written in Go. It runs as a sidecar or local service between an agent and the network, scanning HTTP, WebSocket, and Model Context Protocol traffic through an 11-layer pipeline.
Cost / License
- Freemium (Subscription)
- Open Source (Apache-2.0)
Platforms
- Mac
- Windows
- Linux
- Docker
- Homebrew
- Claude Code
- Cursor
- Visual Studio Code
Features
Pipelock News & Activities
Recent activities
solvionsolutions added Pipelock as alternative to Guardlayer
A-R added Pipelock as alternative to team.management
luckyPipewrench added Pipelock as alternative to AgentGuard
luckyPipewrench added Pipelock as alternative to PromptLock
luckyPipewrench added Pipelock as alternative to Snyk, NemoClaw and AI Security Gateway- luckyPipewrench liked Pipelock
- luckyPipewrench added Pipelock
- POX updated Pipelock
Pipelock information
Developed by
luckyPipewrench Licensing
Open Source (Apache-2.0) and Freemium product.Pricing
Subscription ranging between $49 and $999 per month + free version with limited functionality.Written in
Alternatives
7 alternatives listedSupported Languages
- English
AlternativeTo Categories
AI Tools & Services, Security & PrivacyGitHub repository
- 802 Stars
- 94 Forks
- 3 Open Issues
- Updated
What is Pipelock?
Pipelock is an open-source agent firewall written in Go. It runs as a sidecar or local service between an agent and the network, scanning HTTP, WebSocket, and Model Context Protocol traffic through an 11-layer pipeline. Coverage includes credential exfiltration via 48 DLP patterns, prompt injection detection with 25 patterns and 6-pass normalization, SSRF session binding, and chain detection.
The architecture is capability-separated by design. The agent process holds API keys and credentials and operates without direct network access. Pipelock holds network access and no agent secrets. A compromised proxy cannot leak agent secrets because it never has them.
When signing is enabled, Pipelock emits Ed25519-signed action receipts from outside the agent trust boundary. Each receipt records the action attempted, target, transport, verdict, policy hash, timestamp, and scanner context. A reviewer can verify the receipt chain with the public key, so the evidence does not depend on the agent process attesting to itself.
Distributed as a single ~20MB Go binary. Apache 2.0 license. Listed on the CNCF Landscape under Provisioning > Security & Compliance. Compatible with Claude Code, Cursor, VS Code, JetBrains, OpenAI Agents SDK, Google ADK, AutoGen, CrewAI, and LangGraph.