mandarelabs/mandare

Overview

A TypeScript library that enforces hard spending limits on AI agents, preventing them from exceeding a predefined budget. It works by intercepting agent actions or API calls to stop execution once the budget is exhausted.

Best for

Best for
TypeScript developers building AI agents that need strict, unnegotiable cost controls.

Use cases

• Set a hard cost cap on a single agent run
• Enforce monthly or per-task budgets across multiple agents
• Prevent runaway token usage in production agent workflows

How to use

Install

npx -y @mandarelabs/mcp-server

Tools exposed

• Self-host

Tested with

Claude Code, ChatGPT

Notes

A TypeScript library that enforces hard spending limits on AI agents, preventing them from exceeding a predefined budget. It works by intercepting agent actions or API calls to stop execution once the budget is exhausted.

0 stars on GitHub. Last updated 2026-10-04. Licensed AGPL-3.0.

Use cases

• Set a hard cost cap on a single agent run
• Enforce monthly or per-task budgets across multiple agents
• Prevent runaway token usage in production agent workflows

Pros

• Directly addresses cost overruns with a hard stop
• Simple, focused concept that is easy to reason about
• TypeScript native for developer integration

Cons

• No stars or community traction yet, so maturity is unproven
• Requires integration into existing agent code, which may add overhead
• Limited documentation or examples beyond the tagline

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