Skip to content

Infrastructure for long-running agents

AGNT5 runs your agent and records every step. Crashed runs resume where they stopped, evals tell you if a deploy is better or worse, and bad runs become tests you replay against your next change.

Start for FreeRun Locally

Using a coding agent?

Your agent and harness, defined in code.

Write the agent, the workflow around it, and the scorers that grade it, in Python, TypeScript, or Go. Queues, retries, checkpoints, and timers come with the runtime, so the code you write is only your logic.

Read the build docs
src/support/agent.py
from agnt5 import Agent, Context, tool


@tool
async def lookup_order(ctx: Context, order_id: str) -> dict:
    """Get an order's items, total, delivery date, and final_sale flag."""
    return await orders.get(order_id)  # your order API


@tool
async def search_policy(ctx: Context, question: str) -> str:
    """Search the refund and returns policy."""
    return await policy_docs.search(question)  # your docs index


support_agent = Agent(
    name="support_agent",
    model="anthropic/claude-sonnet-5-5",
    instructions="""You resolve refund requests for Acme.
Look up the order first. Offer a refund only when the
policy allows it, and quote the policy when you do.""",
    tools=[lookup_order, search_policy],
    max_iterations=8,
)

Production, without the platform work.

One command puts support-desk behind an authenticated API, with environments, autoscaling workers, and state kept per customer. You don’t stand any of it up.

Read the run docs
Deploys · deploy 15 · productionlive
$ agnt5 deploy --env production
ok Bundle (1.4s)
ok Deployment created (2.7s)
Replicas active, switching traffic to this deployment...
ok Ready (49.3s) - 2/2 replica(s) active, production routes to this deployment
$ agnt5 deployment checks --env production
After 1 hourDeploy 14Deploy 15
Run failure rate0.4%0.4%no change
New error signaturesnonenoneno change
Latency p506.1s6.0sno change
Cost per run$0.021$0.021no change

The checks run again at 15 minutes and 24 hours.

See every run and make it better.

Every run is traced and measured from its first deploy. Run evals with your scorers on real production runs, and ship the change that tests better.

Read the improve docs
handle_refund · run 7f3aCompleted3d 1h
Input
{"order_id": "A-1042"}
Output
{"item": "Blender", "total": 84.00, "final_sale": false}

Code agent.py:5agent.ts:4agent.go:17

Bring the agents you already have.

No rewrite. Wrap your agent’s entry point in an AGNT5 function and call your framework exactly as you do today. Your prompts, tools, and model providers stay as they are.

  1. Step 1

    Install the CLI

    curl -LsSf https://agnt5.com/cli.sh | bash

    For Python, TypeScript, or Go projects.

  2. Step 2

    Wrap your existing agent

    @function # on your entry point

    One decorator. The agent code inside it doesn’t change.

  3. Step 3

    Deploy

    agnt5 deploy

    Your first run is recorded, start to finish.

Works with

OpenAI Agents SDKClaude Agent SDKGoogle ADKVercel AI SDKCloudflare Workers

Deploy your first agent in 5 minutes.

Start free with 250K steps a month.

Enterprise

Request a demo

Tell us what you're building or how we can help. We'll reply by email.