Getting Started

1Clone the repository

$ git clone https://github.com/Kanevry/agentbus

2Navigate to the project

$ cd agentbus

3Install dependencies

$ npm install

4Run the agent

$ npm start

Or connect to the hosted endpoint: https://agentbus.dev

README

Why

AI agents need events to act on. But connecting webhooks to agents means writing custom HTTP servers, parsing payloads, handling retries, and managing queues — for every single integration.

AgentBus is a lightweight, self-hosted event bus that handles all of that. Receive webhooks. Route to agents. Dispatch actions. Works with any AI framework: CrewAI, LangGraph, OpenAI Agents SDK, or plain HTTP.

Quick Start

npm install agentbus

import { AgentBus } from 'agentbus';

const bus = new AgentBus({
  handlers: [
    {
      name: 'issue-triage',
      pattern: 'github:issues.*',
      handle: async (ctx) => {
        ctx.logger.info(`New issue: ${ctx.event.payload.title}`);
        // Call your AI agent here
        return { success: true };
      },
    },
  ],
});

// Receive an event
await bus.emit({
  source: 'github',
  type: 'issues.opened',
  payload: { title: 'Bug: login fails', labels: ['bug'] },
});

Features

Feature	Community (Free)	Pro
Webhook Ingestion	✅ Unlimited	✅ + Verified Sources
Routing	✅ Glob Patterns	✅ + Content-based
Queue	✅ In-Memory + BullMQ	✅ + Rate Limiting
Storage	✅ SQLite + PostgreSQL	✅ + Event Replay, DLQ
Agents	✅ HTTP Adapter	✅ + A2A, MCP Server
Dispatch	✅ Unlimited	✅ + Dedup, Filters
Support	GitHub Issues	Priority Email

Architecture

Webhooks (GitHub, Stripe, Sentry, ...)
            │
            ▼
     ┌──────────────┐
     │   AgentBus    │
     │  ┌──────────┐ │
     │  │  Router   │ │  ← Glob pattern matching
     │  └──────────┘ │
     │  ┌──────────┐ │
     │  │  Queue    │ │  ← In-memory or BullMQ
     │  └──────────┘ │
     │  ┌──────────┐ │
     │  │  Store    │ │  ← SQLite or PostgreSQL
     │  └──────────┘ │
     └──────────────┘
            │
            ▼
     AI Agents (HTTP, CrewAI, LangGraph, ...)

Integrations

AgentBus works with any AI agent framework:

HTTP — POST events to any endpoint
CrewAI — Trigger crew tasks from webhooks
LangGraph — Send events to LangGraph workflows
OpenAI Agents SDK — Dispatch to OpenAI agents
A2A Protocol — Agent-to-agent communication (Pro)
MCP Server — Model Context Protocol integration (Pro)

Documentation

Visit agentbus.dev for full documentation.

Contributing

See CONTRIBUTING.md for development setup and guidelines.

License

MIT — AgentBus core is free and open source.

@agentbus/pro is available under a commercial license.

Capabilities

StreamingPush NotificationsMulti-TurnAuth: none

View on GitHub

Agentbus