Open Source | MIT License | v0.1.0

The Intent Layer
for AI Agents

A structured, machine-readable graph of your goals, priorities, and intentions — readable and writable by any AI tool.

$ npx goalos init
intent-graph.json
{
  "id": "goal_V1StGXR8",
  "title": "Launch AI Startup",
  "status": "active",
  "priority": {
    "level": "critical",
    "score": 95
  },
  "timeHorizon": "this_quarter",
  "children": 4,
  "progress": 0.35
}
The Problem

Your AI Tools Don't Know What You Want

You use five or more AI tools every day. Each one optimizes in isolation with zero awareness of your bigger picture. More tools should mean more leverage, but instead it means more chaos.

Siloed Intelligence

Each AI tool optimizes in isolation. Your coding agent doesn't know about your Friday pitch. Your writing assistant ignores your product roadmap. They all act like the only tool you use.

Facts Are Not Intentions

Memory tools store facts about you. But knowing you're a data scientist doesn't tell an agent you're launching a startup this month. Facts describe what is. Intent describes what you're trying to achieve.

Context Switching Tax

You re-explain your priorities to every tool, every session. The cognitive overhead grows with each new AI tool you adopt. You're spending more time managing agents than benefiting from them.

The Solution

A Shared Graph of What You're Trying to Achieve

GoalOS gives every AI tool a structured understanding of your goals, priorities, and intentions. Define your intent once. Every connected tool reads from the same source of truth.

Launch Startup
Build Product
Write Manifesto
Ship MCP Server
Get First Users
Publish to npm
Hierarchical Goals Dependency-Aware Permission-Scoped Machine-Readable Event-Driven
How It Works

From Zero to Aligned in Four Steps

Get started in under two minutes. Initialize your intent graph, define your goals, connect your AI tools, and watch them align to what actually matters.

01

Initialize

goalos init
02

Define Goals

graph.addGoal({...})
03

Connect Tools

goalos serve
04

AI Aligns

Every connected tool now shares your context

See It In Action

Developer-First by Design

Clean APIs, strong types, and zero magic. GoalOS works the way developers expect: import a library, call methods, get structured data back.

import { IntentGraph } from '@goalos/core';

// Create your personal intent graph
const graph = IntentGraph.create('abtin', 'Q1 Goals');

// Add a root goal with full context
const startup = graph.addGoal({
  title: 'Launch AI Startup',
  status: 'active',
  priority: { level: 'critical', score: 95 },
  timeHorizon: 'this_quarter',
  motivation: 'Build sovereign AI infrastructure',
  successCriteria: ['10 paying customers', 'Public launch'],
});

// Add a sub-goal with dependency
const landing = graph.addGoal({
  title: 'Ship Landing Page',
  parentId: startup.id,
  status: 'active',
  priority: { level: 'high', score: 80 },
  deadline: '2026-03-15T00:00:00Z',
});

// Every AI tool now reads this context
const priorities = graph.getTopPriorities(5);
const progress = graph.getProgress(startup.id);
from goalos import IntentGraph, Goal, Priority

# Create your personal intent graph
graph = IntentGraph.create("abtin", "Q1 Goals")

# Add a root goal with full context
startup = graph.add_goal(
    title="Launch AI Startup",
    status="active",
    priority=Priority(level="critical", score=95),
    time_horizon="this_quarter",
    motivation="Build sovereign AI infrastructure",
    success_criteria=["10 paying customers", "Public launch"],
)

# Add a sub-goal with dependency
landing = graph.add_goal(
    title="Ship Landing Page",
    parent_id=startup.id,
    status="active",
    priority=Priority(level="high", score=80),
    deadline="2026-03-15T00:00:00Z",
)

# Every AI tool now reads this context
priorities = graph.get_top_priorities(5)
progress = graph.get_progress(startup.id)
# Initialize a new intent graph
$ goalos init --name "Q1 Goals"
  Created ~/.goalos/graph.json

# Add your top-level goal
$ goalos add "Launch AI Startup" \
    --priority critical \
    --horizon this_quarter \
    --domain work
  Added goal_V1StGXR8: Launch AI Startup

# Add sub-goals
$ goalos add "Ship Landing Page" \
    --parent goal_V1StGXR8 \
    --priority high \
    --deadline "2026-03-15"
  Added goal_Kq9mZpL2: Ship Landing Page

# View your goal tree
$ goalos tree
  Launch AI Startup [critical] ........... 35%
    Ship Landing Page [high] ............. active
    Write Manifesto [high] ............... active
    Ship MCP Server [medium] ............. planned

# Start the MCP server for Claude Desktop
$ goalos serve
  GoalOS MCP server running on stdio
Quick Start

Install in Seconds

First-class support for both Python and TypeScript. Pick your language, install the package, and start defining your intent graph.

Py
Python
goalos
$ pip install goalos
from goalos import GoalManager

manager = GoalManager()

# Add a goal with full context
goal = manager.add_goal(
    title="Ship v2.0",
    priority="critical",
    domain="engineering",
    success_criteria=["All tests pass", "Docs updated"]
)

# Get top priorities across all goals
priorities = manager.get_priorities(count=5)
TS
TypeScript
@goalos/core
$ npm install @goalos/core
import { IntentGraph } from '@goalos/core';

const graph = IntentGraph.create('my-project');

// Add a goal with full context
const goal = graph.addGoal({
  title: 'Ship v2.0',
  priority: { level: 'critical', score: 95 },
  domain: 'engineering',
  successCriteria: ['All tests pass', 'Docs updated'],
});

// Get top priorities across all goals
const priorities = graph.getTopPriorities(5);
Ecosystem

Connects to Everything You Use

GoalOS integrates with ChatGPT, Claude, CrewAI, LangChain, and any tool that speaks MCP or OpenAPI. Define your intent once, use it everywhere.

Claude (MCP)

@anthropic/goalos-mcp

Native integration with Claude Desktop and any MCP-compatible client. GoalOS exposes nine tools for the full goal lifecycle: read context, add goals, track progress, and more.

// claude_desktop_config.json
{
  "mcpServers": {
    "goalos": {
      "command": "npx",
      "args": ["@anthropic/goalos-mcp"]
    }
  }
}

ChatGPT (OpenAPI)

spec/openapi.yaml

Use GoalOS with ChatGPT Custom GPTs via the OpenAPI specification. Import the spec as a Custom GPT Action and ChatGPT can read and manage your intent graph directly.

# spec/openapi.yaml
openapi: "3.1.0"
info:
  title: "GoalOS API"
  version: "0.1.0"
paths:
  /goals:
    get:  # List all goals
    post: # Create a new goal
  /goals/{id}/priorities:
    get:  # Get top priorities

CrewAI

goalos.integrations.crewai

Goal-aware CrewAI agents out of the box. The callback automatically feeds intent context into every crew task, so your agents align to what matters most.

from goalos import GoalManager
from goalos.integrations.crewai import GoalOSCrewAICallback

manager = GoalManager()
callback = GoalOSCrewAICallback(manager)

# Your CrewAI agents now read from
# the shared intent graph automatically
crew = Crew(
    agents=[researcher, writer],
    callbacks=[callback]
)

LangChain

goalos.integrations.langchain

Drop-in LangChain callback that injects goal context into chain runs. Every LLM call, tool use, and agent step is aware of what you are trying to achieve.

from goalos import GoalManager
from goalos.integrations.langchain import GoalOSLangChainCallback

manager = GoalManager()
callback = GoalOSLangChainCallback(manager)

# Attach to any LangChain chain or agent
result = chain.invoke(
    {"input": "Plan next sprint"},
    config={"callbacks": [callback]}
)
Use Cases

Built for How You Actually Work

Whether you're an individual power user or leading a team, GoalOS adapts to your workflow. Define intent at any scale.

AI Power Users

Connect five or more AI tools through a single intent graph. Every tool you add makes every other tool more useful. Stop re-explaining your priorities and start compounding your leverage.

Agent Developers

Build agents that understand user context from day one. Read the intent graph to know what matters, respect permissions, and ship AI that's aligned by design rather than by accident.

Startup Founders

One intent graph connects your coding agent, writing assistant, and project manager. Your Friday pitch prep informs your Monday sprint. Context flows across every tool you touch.

Engineering Teams

Shared intent graphs for sprint goals, project milestones, and team alignment. Scoped permissions let each agent access only what it needs. Everyone works from the same source of truth.

Platform

Complete Developer Toolkit

Everything you need to integrate structured intent into your AI stack. TypeScript core, Python SDK, CLI, MCP server, and a spec that's extensible by design.

@goalos/core

TypeScript core library with full type safety, schema validation, event system, dependency resolution, and merge strategies. The foundation everything else builds on.

@goalos/mcp-server

Model Context Protocol server for Claude Desktop and any MCP-compatible client. Nine tools covering the complete lifecycle from reading context to completing goals.

goalos (CLI)

Command-line tool for managing intent graphs. Initialize, add goals, set priorities, view tree structures, export data, and start the MCP server from your terminal.

goalos (Python)

Full Python SDK with Pydantic models, async support, and an API that mirrors the TypeScript library exactly. First-class support for the Python AI ecosystem.

JSON-LD Format

Linked Data serialization for interoperability and semantic web compatibility. Intent graphs are portable, self-describing, and ready for any system that speaks JSON.

Permission System

Granular agent permissions: read, write, complete, create sub-goals, and reprioritize. Scope access by goal, domain, or tree depth. Your intent, your rules.

0
MCP Tools
0
Goal Operations
0
Example Graphs
0
SDKs
Works With
Claude Desktop LangChain CrewAI AutoGPT Any MCP Client

Ready to Align Your AI Stack?

Open source. Self-hosted. Your data stays yours. Define your intent once and let every AI tool work toward what actually matters.

$ npm install @goalos/core