Quick Start

This guide creates a reusable agent and environment, starts a session, and prints the agent’s response to its first message.

Prerequisites

You need:

A Funky API key
Python 3.10 or newer, or Node.js 20 or newer

1. Install the SDK

Python
TypeScript

pip install funky-sdk

npm install funky-sdk

2. Set your API key

The SDK reads your API key from the FUNKY_API_KEY environment variable.

export FUNKY_API_KEY="fk_..."

Keep your API key out of source control. In production, load it from your platform’s secret manager.

3. Create your first session

Create quickstart.py for Python or quickstart.ts for TypeScript. Add the snippet from each step to the file, in order.

Create an agent

An agent defines the model, system prompt, and tool policy. You can reuse the same agent across many sessions.

Python
TypeScript

from funky import Funky

funky = Funky()

agent = funky.agents.create(
    name="Repository investigator",
    system_prompt=(
        "You are a careful coding agent. "
        "Verify claims with tools."
    ),
    model={
        "provider": "anthropic",
        "model": "claude-sonnet-5",
    },
    tool_policy={"max_iterations": 20},
)

import { Funky } from "funky-sdk";

const funky = new Funky();

const agent = await funky.agents.create({
  name: "Repository investigator",
  system_prompt: "You are a careful coding agent. Verify claims with tools.",
  model: {
    provider: "anthropic",
    model: "claude-sonnet-5",
  },
  tool_policy: { max_iterations: 20 },
});

Create an environment

An environment configures the sandbox where the agent runs. This example limits outbound network access to GitHub.

Python
TypeScript

environment = funky.environments.create(
    name="github-access",
    network={
        "type": "limited",
        "allowed_hosts": [
            "github.com",
            "*.githubusercontent.com",
        ],
    },
)

const environment = await funky.environments.create({
  name: "github-access",
  network: {
    type: "limited",
    allowed_hosts: [
      "github.com",
      "*.githubusercontent.com",
    ],
  },
});

Start a session

A session connects the agent to the environment for one independent task or conversation. Wait until its sandbox is ready before sending work.

Python
TypeScript

session = funky.sessions.create(
    agent=agent.id,
    environment_id=environment.id,
    title="Investigate an issue",
)

funky.sessions.wait_until_ready(session.id, timeout=180)

const session = await funky.sessions.create({
  agent: agent.id,
  environment_id: environment.id,
  title: "Investigate an issue",
});

await funky.sessions.waitUntilReady(session.id);

Send a message

run_turn() and runTurn() send the message, follow session events until the turn completes, and return the agent’s response.

Python
TypeScript

result = funky.sessions.run_turn(
    session.id,
    content=(
        "Inspect the repository and identify the root cause "
        "of the reported issue."
    ),
)

print(result.output_text)
funky.close()

Run the completed file:

python quickstart.py

const result = await funky.sessions.runTurn(session.id, {
  content:
    "Inspect the repository and identify the root cause of the reported issue.",
});

console.log(result.output_text);
funky.close();

Run the completed file with your project’s TypeScript runner. For example:

npm install --save-dev tsx
npx tsx quickstart.ts

In an application, create agents and environments during setup and store their IDs. Reuse those IDs when you start a new session for each independent task.

Getting started

Agent Config

Environment Config

Session

Prerequisites

1. Install the SDK

2. Set your API key

3. Create your first session

​Prerequisites

​1. Install the SDK

​2. Set your API key

​3. Create your first session

Prerequisites

1. Install the SDK

2. Set your API key

3. Create your first session