Quick Start

This guide self-hosts V1 — the free, open-source, single-user edition that works for you (goals, memory, documents, messaging delivery). V2 adds working as you — human proxy across email, meetings & phone, plus multi-user login, admin console, and more. V2 is invite-only and available now; it is not part of the public repo, so you can't self-host it. If you're on the hosted site, you're already using V2. See V1 vs V2.

Prerequisites

Docker Desktop

Must be running on your machine. Download here.

One LLM API key

OpenRouter is the fastest to set up — free tier available, 100+ models, one key. Sign up at openrouter.ai.

No GPU required. AI Partner runs entirely on cloud LLMs. Ollama (local models) is optional and can be added later.

Step 1: Clone and start

git clone https://github.com/AmitkrPaliwal/AI-Partner
cd AI-Partner
cp .env.example .env
docker compose up --build

The first build takes 2–4 minutes while Docker pulls images and installs dependencies. When you see:

[app] Server running on :3000
[app] Socket.IO ready

...you're ready.

Step 2: Open the UI

Open http://localhost:3000 in your browser.

On first run, the Setup Wizard walks you through:

Choosing a primary LLM provider
Entering your first API key
Setting your name and role

You can skip the wizard at any time and configure everything from Settings → Models.

Step 3: Add your LLM key

Go to Settings → Models → Add Provider and paste your key:

Provider	Where to get it	Notes
OpenRouter	openrouter.ai	Free tier, 100+ models — easiest to start
Anthropic	console.anthropic.com	Claude 3.5 / 4 models
OpenAI	platform.openai.com	GPT-4o and others
Groq	console.groq.com	Free tier, very fast (~100k tokens/s)

You can also open .env in any text editor, set OPENROUTER_API_KEY=your_key_here, and run docker compose restart app.

Step 4: Run your first goal

Click the chat input and type this exactly:

Research the 5 most-starred GitHub repositories in the last 7 days.
Write a summary with each repo's name, stars, language, and a one-line description.
Then generate an Excel file with the results.

Press Enter or click Send as Goal (the rocket icon).

Watch the Goal Progress panel on the right:

Decomposing — the goal is split into sub-tasks (research, extract, generate file)
Executing — you'll see each tool call in real time: web_search → web_fetch → execute_python → generate_excel
Validating — the executor checks that the file was created and contains data
Done — a download link appears in the Files panel

The whole flow takes 30–90 seconds on the first run.

If the goal panel doesn't appear, make sure you clicked Send as Goal (rocket icon), not the plain chat send. The chat mode is for quick Q&A; the goal mode runs the full ReAct loop.

Step 5: Configure your identity

For the full experience — proactive tasks, meeting attendance, email proxy — AI Partner needs to know who you are. Edit these four files in Settings → Workspace (or directly in the workspace/ folder):

USER.md — who you are

This is injected into every prompt the agent sees. Keep it under 2,000 characters.

# User Profile
Name: Alex Chen
Role: Founder & CEO, Acme Inc.
Current focus: Closing Series A, Q2 2026
Key relationships: Sarah (CFO), Mike (CTO), Priya at Sequoia
Communication style: Direct. No filler. Lead with data.

GOALS.md — what you're working toward

Meeting attendance, proactive tasks, and heartbeat briefings all align to these goals automatically.

## Active Goals

### Close Series A
- Success: Term sheet signed by June 30, 2026
- Context: In conversations with Sequoia, Accel, and Tiger Global

### Launch v2.0
- Success: 1,000 active users by July 15, 2026
- Context: Feature freeze April 30; beta starts May 15

HEARTBEAT.md — what to monitor automatically

These run on your schedule without any prompt from you.

## Standing Tasks

- Every morning (7 AM): Summarize top 5 tech and startup news stories
- Every weekday (9:15 AM): Check NIFTY 50 and my portfolio stocks
- Every Friday (5 PM): Summarize this week's meeting transcripts and key decisions

AUTHORITY.md — what it can do vs. what needs your approval

This is the most important file. It determines which actions execute automatically and which are sent to you via Telegram for a one-tap approval.

## Auto (execute immediately, audit-logged)
- action: read_.*
- action: send_message + counterparty: family

## Draft + ask (send me a Telegram approval request)
- action: reply_email + counterparty: colleague
- action: reply_email + counterparty: client
- action: send_slack_dm + counterparty: colleague

## Block (never do, period)
- action: payments.*
- action: legal_commitment
- action: place_phone_call