Quick Start

Quick Start🔗

This guide will get you chatting with LocalGPT in just a few minutes.

This is the assistant’s quick start. For worlds, see the desktop app and LocalGPT Gen; to previs a screenplay with nothing installed, drop it on localgpt.md.

1. Configure a Model Provider🔗

Choose one of the following options:

Option A: LM Studio (Local, No API Key)🔗

Run models entirely on your machine with LM Studio:

# Install LM Studio from https://lmstudio.ai
# Download a model and start the local server, then:
localgpt config set agent.default_model "openai/qwen/qwen3.5-35b-a3b"
localgpt config set providers.openai.api_key "lm-studio"
localgpt config set providers.openai.base_url "http://127.0.0.1:1234/v1"

Option B: Ollama (Local)🔗

# Install Ollama from https://ollama.ai, then:
ollama pull llama3
localgpt config set agent.default_model "llama3"

Option C: Claude CLI🔗

If you have the Claude CLI installed and authenticated:

# No additional setup needed! LocalGPT uses your Claude CLI auth.
# Just start chatting with:
localgpt chat

Option D: API Keys🔗

# For OpenAI
export OPENAI_API_KEY="sk-..."

# For Anthropic API
export ANTHROPIC_API_KEY="sk-ant-..."

# For Gemini
export GEMINI_API_KEY="..."

2. Start an Interactive Chat🔗

localgpt chat

You’ll see a prompt where you can type messages:

LocalGPT Chat (type /help for commands, /quit to exit)
> Hello! What can you help me with?

I can help you with a variety of tasks:
- Answer questions and explain concepts
- Write and review code
- Execute shell commands
- Search and manage your memory
- And much more!

>

3. Basic Chat Commands🔗

While in chat, use these commands:

Command	Description
/help	Show available commands
/quit	Exit the chat
/memory <query>	Search your memory
/save	Force save important context
/compact	Manually compact the session
/status	Show session status
/clear	Clear the screen

4. Ask a Single Question🔗

For quick one-off questions:

localgpt ask "How do I create a new git branch?"

5. Search Your Memory🔗

As you chat, important information is automatically saved. Search it later:

# Search memory from CLI
localgpt memory search "git commands"

# Or from within chat
> /memory git commands

6. Start the Daemon🔗

For the HTTP API and heartbeat functionality:

# Start in foreground
localgpt daemon start

# Check status
localgpt daemon status

# Stop the daemon
localgpt daemon stop

Example Session🔗

$ localgpt chat

LocalGPT Chat (type /help for commands, /quit to exit)

> Can you help me write a Python script that reads a CSV file?

Sure! Here's a simple Python script using the csv module:

```python
import csv

with open('data.csv', 'r') as file:
    reader = csv.DictReader(file)
    for row in reader:
        print(row)

Save this to a file called read_csv.py

I’ll create that file for you. [Tool: write_file] Creating read_csv.py…

Done! I’ve created read_csv.py with the CSV reading code.

/memory csv Found 1 result for “csv”:

• [2026-02-14] Discussed CSV file reading in Python

/quit Goodbye!

## Next Steps

- Learn about [CLI Commands](/docs/cli-commands)
- Understand the [Memory System](/docs/memory-system)
- Set up [Skills](/docs/skills) for specialized tasks
- Configure [Heartbeat](/docs/heartbeat) for autonomous tasks