Stop runaway coding agents before they spend.
AgentGuard47 is a zero-dependency Python SDK that enforces budgets, loops, retries, and timeouts while your agent is still running. No account, API key, or dashboard required for the first proof.
pip install agentguard47
Give the skeptical developer proof before asking for anything.
The homepage points non-signup users at a concrete local sequence they can copy into a clean Python environment and inspect without talking to us.
agentguard doctor
agentguard demo
agentguard quickstart --framework raw --write
Four failure modes, one small SDK.
Budget burn
Set call, token, or dollar ceilings before a coding agent turns a bad loop into a bill.
Tool loops
Stop repeated tool calls and repeated argument paths while the run is still live.
Retry storms
Cap retry attempts when a tool, provider, or repo step keeps failing the same way.
Timeout drift
Make long-running agent work fail clearly instead of hanging until the operator notices.
Start local. Add hosted control only when the work becomes shared.
Best first action
Run `doctor`, `demo`, and a generated quickstart file locally.
No account, no dashboard, no network requirement.
When it matters
Use retained incidents, alerts, team visibility, and remote kill.
Useful after a workflow has real runs worth operating.
Boundary
AgentGuard47 focuses on runtime enforcement and incident control.
It is not prompt management, eval orchestration, or a trace warehouse.
The real integration is small.
Plain Python
import agentguard
tracer = agentguard.init(
service="repo-coder",
budget_usd=5.00,
trace_file="traces.jsonl",
local_only=True,
)
with tracer.trace("agent.run") as span:
span.event("tool.call", data={"tool": "edit"})
After local proof
Open the quickstart for OpenAI, Anthropic, LangChain, LangGraph, CrewAI, or raw Python. Then connect the hosted dashboard only when the trace needs to be retained or shared.
Run the local proof. Keep the dashboard optional.
Developers who are not ready to sign up still leave with a working install path, a trace file, and a guard they can copy into a real coding-agent loop.