Forge Agent

The open agent runtime.

Deploy AI agents that bridge every platform. WhatsApp, Telegram, Discord, Slack, CLI -- one config, every channel.

Get Started → See How It Works
How It Works

From config to deployment in four steps

Define your agent in YAML. Forge handles the routing, bridging, and process management automatically.

Step 01

YAML Config

Define your agent's personality, channels, skills, and LLM preferences in a single YAML file. No boilerplate code required.

Step 02

LLM Routing

Forge routes each request to the optimal model. Claude for reasoning, GPT for conversation, DeepSeek for code. Automatic failover included.

Step 03

Multi-Channel Bridges

One agent instance connects to every platform simultaneously. WhatsApp, Telegram, Discord, Slack, and CLI all share the same memory and context.

Step 04

MCP Skills

Extend your agent with MCP-compatible skills. Database queries, API calls, file operations, web search -- plug in any capability.

Features

Everything you need to ship production agents

Six core capabilities that handle the hard parts of building, deploying, and managing AI agents at scale.

Multi-Channel Bridges

Connect to WhatsApp, Telegram, Discord, Slack, and CLI from a single agent instance. Add new channels with one line of config.

LLM Router

Intelligent routing across Claude, GPT, DeepSeek, and Gemini. Set primary and fallback models per agent. Automatic failover on rate limits or downtime.

MCP Skills

Extend agents with Model Context Protocol skills. Database access, file I/O, web search, API integrations -- any MCP-compatible tool works out of the box.

SQLite Memory

Persistent conversation memory stored in SQLite. Agents remember context across sessions and channels. No external database required.

WebSocket Gateway

Real-time communication layer with heartbeat monitoring and auto-reconnection. Build live dashboards and monitoring interfaces on top of the agent runtime.

PM2 Process Management

Production-grade process management with automatic restarts, log rotation, and cluster mode. Deploy once and forget about uptime.

Architecture

One runtime, every channel

Forge Agent sits between your LLM providers and your communication channels, handling routing, memory, and skills in a single process.

Config
agent.yaml
Core
Forge Agent
Gateway
WebSocket
WhatsApp
Telegram
Discord
Slack
CLI
MIT Licensed

Open source. Built in public.

Forge Agent is fully open source under the MIT license. Star us on GitHub, open a PR, or fork it and make it yours.

Star on GitHub →

Ship your first agent today.

Get Started → Talk to Us