jamjet-labs/jamjet

Overview

An open-source safety layer for AI agents. It blocks unsafe tool calls, requires approval, enforces budgets, and provides audit and replay capabilities.

Best for

Best for
Teams building AI agents that need safety and audit controls without vendor lock-in

Use cases

• Enforcing spending limits on agent tool usage
• Requiring human approval for sensitive tool calls
• Auditing and replaying agent actions for compliance

How to use

Install

pip install jamjet

Tested with

Claude Desktop, Claude Code, Cursor, ChatGPT, OpenAI Agents SDK, LangChain, CrewAI, ADK

Example client config

# ~/.jamjet/policy.yaml\nversion: 1\nrules:\n  - { match: "*delete*", action: block }\n  - { match: "payments.*", action: require_approval }\n  - { match: "shell.exec", action: block }

Notes

An open-source safety layer for AI agents. It blocks unsafe tool calls, requires approval, enforces budgets, and provides audit and replay capabilities.

13 stars on GitHub. Last updated 2026-06-01. Licensed Apache-2.0.

Use cases

• Enforcing spending limits on agent tool usage
• Requiring human approval for sensitive tool calls
• Auditing and replaying agent actions for compliance

Pros

• Open-source and self-hostable, no vendor lock-in
• Built in Rust for performance and reliability
• Provides multiple safety controls in one tool: approval, budgets, audit, replay

Cons

• Early stage project with small community (13 stars)
• Adds latency and overhead to agent workflows
• Requires integration effort with existing agent frameworks

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.