pi-mono: TypeScript Full-Stack AI Agent Toolkit

Why pi-mono Exists

The dominant AI agent tooling is Python: LangChain, LlamaIndex, Dify, CrewAI. But if you’re building products in the TypeScript / Node.js ecosystem and want AI capabilities, the options are limited.

Most teams end up with:

• Backend: Go / Node.js / Java
• Frontend: React / Flutter
• AI integration: either direct OpenAI API calls or a Python microservice layer

pi-mono tries to solve this. It’s a TypeScript monorepo offering a complete toolkit from low-level LLM API to high-level coding agent.

Architecture Overview

pi-mono/
├── packages/
│   ├── ai/              # Unified LLM API
│   ├── agent/           # Agent runtime + tool calling
│   ├── coding-agent/    # Interactive coding agent CLI
│   ├── tui/             # Terminal UI library
│   ├── web-ui/          # Web chat components
│   ├── mom/             # Slack bot
│   └── pods/            # vLLM deployment CLI

Core Components

1. @mariozechner/pi-ai — Unified Multi-Provider LLM API

import { createAI } from "@mariozechner/pi-ai";

// One interface, switch providers
const ai = createAI({
  provider: "openai", // or "anthropic", "google"
  model: "gpt-4o",
  apiKey: process.env.OPENAI_API_KEY
});

// Unified API regardless of provider
const response = await ai.complete("Hello, world");

Supported providers:

• OpenAI (GPT-4o, GPT-4o-mini, etc.)
• Anthropic (Claude 3.5 Sonnet, Opus, etc.)
• Google (Gemini 1.5, etc.)

Value: If your product needs to switch between models (cost optimization, regional compliance), a unified API layer reduces lock-in.

2. @mariozechner/pi-agent-core — Agent Runtime

This is the core. pi-agent-core implements:

• Tool calling: Give agents ability to call tools
• State management: Cross-turn state persistence
• Streaming: Streaming output support

import { createAgent } from "@mariozechner/pi-agent-core";

const agent = createAgent({
  ai: openAIInstance,
  tools: [
    {
      name: "read_file",
      description: "Read file contents",
      parameters: {
        path: "string"
      },
      execute: async ({ path }) => {
        return readFileSync(path, "utf-8");
      }
    }
  ]
});

// Agent automatically decides when to call tools
const result = await agent.run("Read package.json to check dependencies");

3. @mariozechner/pi-coding-agent — Interactive Coding Agent CLI

A developer tool. Install it and chat with AI directly in your terminal for code-related tasks.

# Install
npm install -g @mariozechner/pi-coding-agent

# Run
pi-agent

# Then chat
> add dark mode support to this React component

4. @mariozechner/pi-pods — vLLM Deployment CLI

For self-hosting open-source models, pi-pods manages vLLM deployments on GPU pods via CLI.

# Deploy a vLLM instance
pi-pods deploy --model mistralai/Mistral-7B-Instruct-v0.2

# Check status
pi-pods list

Use case: running open-source models on your own GPUs with an API interface.

Comparison with LangChain.js

Practical Use Cases

Scenario 1: Add AI to Node.js Backend

Express service needing AI to analyze user input:

import { createAI } from "@mariozechner/pi-ai";
import { createAgent } from "@mariozechner/pi-agent-core";

const ai = createAI({ provider: "anthropic", model: "claude-3-5-sonnet" });
const agent = createAgent({ 
  ai,
  tools: [yourTools]
});

app.post("/analyze", async (req, res) => {
  const result = await agent.run(req.body.userInput);
  res.json({ result });
});

Scenario 2: Internal Coding Agent

Set up a coding agent CLI for your team to manipulate codebases via natural language:

# Developer runs locally
pi-agent

> add unit tests for the payment module
> refactor this function to use async/await
> explain what this regex does

Limitations

• Not a visual platform like Dify/CrewAI: pi-mono is a codebase—you write TypeScript to integrate
• Provider coverage limited: Doesn’t cover all LLM providers (no Cohere, Azure OpenAI, etc.)
• Production validation: Less production use cases compared to LangChain

Installation

# Full install
npm install @mariozechner/pi-ai @mariozechner/pi-agent-core ...

# Or per-package
npm install @mariozechner/pi-ai

Requires Node.js 20+.

Conclusion

pi-mono’s value proposition: TypeScript-native + full-stack coverage + modular packages.

If you’re in the TypeScript ecosystem and want to add AI capabilities without a Python microservice, pi-mono is worth exploring. Its modular design lets you use just pi-ai for API unification, or go full-stack with a complete coding agent setup.

Repo: github.com/badlogic/pi-mono