aggrete/aggrete

Overview

Aggrete is a policy proxy for AI assistants. It sits between an assistant and upstream tools, refusing forbidden tool calls before they reach the target. Written in Python, it governs what assistants can reach and do.

Best for

Best for
Developers who need a lightweight policy gate for AI assistant tool calls

Use cases

• Enforce allow or deny policies on AI tool calls
• Block disallowed actions before upstream execution
• Add a governance layer for assistant tool access

How to use

Install

uvx aggrete --demo                  # the walkthrough — no config, auth, or network

Tested with

Claude Code

Notes

Aggrete is a policy proxy for AI assistants. It sits between an assistant and upstream tools, refusing forbidden tool calls before they reach the target. Written in Python, it governs what assistants can reach and do.

1 stars on GitHub. Last updated 2026-09-15. Licensed Apache-2.0.

Use cases

• Enforce allow or deny policies on AI tool calls
• Block disallowed actions before upstream execution
• Add a governance layer for assistant tool access

Pros

• Simple Python implementation
• Prevents forbidden calls before upstream contact
• Lightweight proxy approach

Cons

• Very early stage with minimal adoption
• Limited documentation or community
• Python only

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