agenthelm-sdk 0.1.0

Python SDK for AgentHelm — AI Agent Control Plane

agenthelm-sdk

Monitor your AI agents from AgentHelm. Add one line. See everything.

Install

pip install agenthelm-sdk

Quick Start

import agenthelm

Connect your agent (one line):

dock = agenthelm.connect( "ahe_live_xxxxx", name="My Agent" )

Send logs:

dock.log("Agent started") dock.log("Warning: rate limit", level="warning") dock.log("Task complete", level="success")

Track token usage:

dock.track_tokens(used=1500, model="gemini-flash") dock.track_tokens( used=2000, model="gpt-4", cost_per_1k=0.03 )

Send structured output:

dock.output({ "leads_found": 12, "hot_leads": 5, "report_sent": True })

Report errors:

try: risky_operation() except Exception as e: dock.error("Operation failed", exception=e)

Handle dashboard commands:

@dock.on_command("stop") def handle_stop(payload): dock.log("Stopping agent...") dock.stop()

@dock.on_command("run_now") def handle_run(payload): # run_main_task() dock.reply("Task started!")

Chat with your agent:

@dock.on_chat def handle_chat(message: str): if "status" in message.lower(): dock.reply("Running fine!") else: dock.reply(f"Received: {message}")

Keep agent alive:

dock.listen()

Context Manager

with agenthelm.connect("ahe_live_xxxxx") as dock: dock.log("Working...") # do_work() dock.output({"done": True})

Agent auto-stops when block exits.

Configuration

dock = agenthelm.connect( key="ahe_live_xxxxx", name="My Agent", agent_type="python", version="2.1.0", verbose=True, ping_interval=30, command_poll_interval=5, timeout=5 )

Links

Dashboard: https://agenthelm.dev Docs: https://agenthelm.dev/docs Support: [e-mail]