funky

Agent swarms on demand

Spin up hundreds of sub-agents instantly
without thinking about the infrastructure.

How it works

Hand off work to a swarm of sub-agents in parallel.
Collect results back when it's done.

1

Parent agent defines work

2

Fans out to sub-agents

3

Results returns back

import os
from funky import Funky

client = Funky(api_key=os.getenv("FUNKY_API_KEY"))
sub_agent = client.create_subagent(model="gemini-3-flash-preview")
try:
    messages = sub_agent.send_message("What is best restaurant in SF?")
    print(messages[-1].text)
finally:
    sub_agent.terminate()

Use cases

One API, various use cases.

Coding agents

Distribute code generation, testing, and refactoring across many parallel agents. Ship faster without burning tokens on a single expensive model.

Research agents

Fan out to hundreds of sources simultaneously. Synthesize findings in seconds instead of hours. Perfect for due diligence and market analysis.

Browser agents

Crawl, scrape, and interact with the web at scale. Each sub-agent gets its own browser session with full isolation.

Every sub-agent runs in its own sandbox.

Isolated execution environments. No cross-contamination. Full resource control.

sub-agent-01

running

sub-agent-02

running

sub-agent-03

running

sub-agent-04

running

sub-agent-05

running

Your sub-agents work fine,
until you need a hundred of them.

10,000+

sub-agents spawned this month

100

sub-agents spawned in 1 second

Spawn your first sub-agent