Quick Start

Quick Start

Get Nebulus Atom running and working on a task in 5 minutes.

1. Start Your LLM Server

Make sure your local LLM server is running. For example, with Ollama:

ollama serve
ollama pull llama3.1:8b

2. Configure and Launch

cd nebulus-atom
source .venv/bin/activate

# Set your LLM server URL (adjust for your server)
export NEBULUS_BASE_URL=http://localhost:11434/v1
export NEBULUS_MODEL=llama3.1:8b

# Start the agent
python3 -m nebulus_atom.main start

3. Try Some Commands

Once the agent is running, try these:

> list files in the current directory
> read the README.md file
> create a Python function that calculates fibonacci numbers
> /help

4. Use Context Pinning

Pin important files so the agent always has them in context:

> pin src/main.py
> pin tests/test_main.py
> show context

5. Run with an Initial Prompt

You can pass a prompt directly at startup:

python3 -m nebulus_atom.main start "fix the bug in auth.py"

6. Use Session IDs

Separate conversations with session IDs:

python3 -m nebulus_atom.main start --session-id feature-auth
python3 -m nebulus_atom.main start --session-id bugfix-login

7. Launch the Dashboard

View agent telemetry in the Flight Recorder dashboard:

python3 -m nebulus_atom.main dashboard

Opens a Streamlit dashboard at http://localhost:8501.

Key Concepts

Concept	Description
Context	Files pinned to the conversation for the agent to reference
Checkpoints	Auto-saved snapshots before risky file operations
Skills	Reusable capabilities the agent can learn and execute
RAG	Semantic code search using embeddings
TDD Mode	Test-driven development cycle automation
Auto Mode	Autonomous execution without confirmation prompts

What's Next?

• CLI Reference - Full command documentation
• Features - All available features
• Nebulus Swarm - Multi-agent orchestration