AgentForge

Overview

Developer tool for creating production-ready AI agents from YAML configs. Visual dashboard with voice and text channels, LangSmith observability.

AgentForge solves the problem of rebuilding AI agent infrastructure from scratch for every project. Define an agent's personality, tools, memory, and routing in a YAML config, and the platform generates a production-ready agent with voice and text channels. The Next.js dashboard provides a visual editor, prompt playground, and evaluation runner. The Python runtime handles LLM routing across OpenAI, Anthropic, and Google, with automatic LangSmith tracing for observability. A knowledge system supports markdown files with wikilinks for graph-based retrieval.

Category

AI

Stack

Next.jsPythonFastAPILangSmithLiveKitSupabase

Features

Visual Dashboard

Agent management, YAML config editor with live validation, Monaco-powered BAML prompt editor, prompt playground, and evaluation runner.

Voice Channel

LiveKit-powered real-time voice with Deepgram STT, Cartesia TTS, and Silero VAD. Test voice agents directly from the dashboard.

Web Chat Channel

HTTP and WebSocket chat interface. Embeddable widget for testing and production deployment.

Knowledge System

Markdown files with YAML frontmatter and wikilinks build a knowledge graph. Keyword + graph-based retrieval with quality gates.

LangSmith Tracing

Automatic trace capture for every agent interaction. Built-in trace viewer and evaluation framework for latency, tokens, task completion, and safety.

YAML Config

Portable YAML configurations define agent personality, tools, memory, and routing. Hot reload on config changes during development.

Architecture

Dashboard

Next.js 14TypeScriptshadcn/uiTailwindZustandMonaco EditorReact Flow

Runtime

FastAPIPythonOpenAIAnthropicLiveKitLangSmithBAML

Infrastructure

SupabasePostgreSQLDockerTurborepopnpm

Workflow

1 Write YAML config
2 Test in playground
3 Evaluate with LangSmith
4 Deploy agent