An open-source framework for creating agents that discover tasks, execute them, and self-heal. 10 lines of code. Zero dependencies.
Here is a complete uptime monitor agent. It discovers URLs to check, pings them on a schedule, and reports results — all in one file.
import { createAgent } from "@zoobicon/agents";
const monitor = createAgent({
id: "uptime-check",
name: "Uptime Monitor",
scheduleIntervalSec: 300,
discover: async () => [
{ url: "https://mysite.com" },
{ url: "https://api.mysite.com/health" },
],
execute: async (input) => {
const res = await fetch(input.url);
return {
output: { status: res.status, up: res.ok },
confidence: 1,
};
},
});
await monitor.run();Batteries included, but every piece is replaceable.
Agents find their own work. Define what to look for, the framework handles scheduling and deduplication.
Every execution returns a confidence score. Auto-execute when confidence is high, flag for review when low.
Exponential backoff, configurable retries. Agents recover from failures automatically without intervention.
InMemory for testing, PostgreSQL for production. Bring your own storage backend with a simple interface.
Extend agents with community-built skills. Like npm for AI capabilities. Install, compose, and share.
Subscribe to agent events. Build dashboards, alerts, and integrations with hooks for every lifecycle stage.
A simple loop: discover, execute, store, emit. Failures retry automatically.
discover()
│
▼
[tasks] ──────► execute(task)
│
▼
┌─── success? ───┐
│ │
yes no
│ │
▼ ▼
{ output, retry with
confidence } backoff
│ │
▼ └──► (max retries → fail event)
store results
│
▼
emit events
(dashboard, alerts, hooks)Zoobicon Agents is built for the serverless era — minimal surface area, maximum flexibility.
| Feature | Zoobicon Agents | OpenClaw | AutoGPT |
|---|---|---|---|
| Setup time | 2 minutes | 30+ minutes | 1+ hours |
| Runtime | Serverless (Vercel/AWS) | Local process | Local process |
| Language | TypeScript | TypeScript | Python |
| Dependencies | 0 (core) | Many | Many |
| Storage | Pluggable (Postgres/Memory) | Markdown files | Vector DB |
| License | MIT | MIT | MIT |
| GitHub Stars | New | 163K | 170K |
| Enterprise hosted | Yes (Zoobicon) | No | No |
Zoobicon Agents Cloud runs your agents 24/7 with managed cron, dashboards, alerts, and team collaboration.
From zero to a running agent in under five minutes.
npm install @zoobicon/agentsimport { createAgent } from "@zoobicon/agents";
const agent = createAgent({
id: "my-agent",
name: "My First Agent",
discover: async () => [{ task: "hello" }],
execute: async (input) => ({
output: { message: "Hello from agent!" },
confidence: 1,
}),
});await agent.run();
// Agent discovers tasks and executes themconst agent = createAgent({
id: "scheduled-agent",
name: "Scheduled Agent",
scheduleIntervalSec: 300, // every 5 minutes
discover: async () => [ /* ... */ ],
execute: async (input) => { /* ... */ },
});import { PostgresAgentStore } from "@zoobicon/agents/stores";
const store = new PostgresAgentStore(process.env.DATABASE_URL);
const agent = createAgent({
id: "persistent-agent",
name: "Persistent Agent",
store, // results survive restarts
discover: async () => [ /* ... */ ],
execute: async (input) => { /* ... */ },
});# Works on Vercel, AWS Lambda, any Node.js server
vercel deploy
# Or run as a long-lived process
node agent.mjs