bmdhodl/agent47

Overview

AgentGuard is a Python library that enforces runtime cost guardrails for AI agents. It provides budget enforcement, loop detection, and a kill switch to prevent runaway spending. The tool has zero dependencies and is MIT licensed.

Best for

Best for
Developers building AI agents who need a lightweight, no-frills cost guardrail

Use cases

• Set a hard budget cap on agent API calls to avoid cost overruns
• Detect and halt infinite loops in agent execution
• Implement a kill switch to manually stop an agent mid-task

How to use

Install

pip install agentguard47

Tested with

Claude Code, Cursor, Cline, ChatGPT

Example client config

from agentguard import BudgetGuard, LoopGuard, RateLimitGuard, Tracer\nbudget = BudgetGuard(max_calls=20, max_cost_usd=1.00)\nloop = LoopGuard(max_repeats=2)\nrate = RateLimitGuard(max_calls_per_minute=10)\ntracer = Tracer(service="cursor-agent", guards=[loop, rate])\nwith tracer.trace("agent.run"):\n    budget.consume(calls=1)\n    # tool call here — guards raise on overrun

Notes

AgentGuard is a Python library that enforces runtime cost guardrails for AI agents. It provides budget enforcement, loop detection, and a kill switch to prevent runaway spending. The tool has zero dependencies and is MIT licensed.

3 stars on GitHub. Last updated 2026-06-07. Licensed MIT.

Use cases

• Set a hard budget cap on agent API calls to avoid cost overruns
• Detect and halt infinite loops in agent execution
• Implement a kill switch to manually stop an agent mid-task

Pros

• Zero dependencies makes it easy to integrate
• MIT license allows unrestricted use and modification
• Simple, focused solution for a common cost problem

Cons

• Limited to Python environments only
• No built-in monitoring or alerting beyond the kill switch
• Small community and limited documentation due to low stars

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