Axion Labs Documentation

Lumen & Veil

Free to chat with after creating an Axion account. Browser chat does not require an API key.

Free account required

Lumen 1.2.5

model: "lumen"

Fine-tuned Llama 3.1 8B with DPO safety training. General assistant — questions, writing, code, explanations. Also available at axion.amplifiedsmp.org/lumen.

Deprecated

Veil 1.1

model: "veil"

Our first model — a fine-tuned Llama 3.1 3B for quick general chat. Deprecated and on its way out; see why we\'re moving on.

Veil is deprecated. Don\'t build anything new against model: "veil" — point it at model: "lumen" on the same endpoint instead. Existing integrations should migrate now rather than waiting for the retirement date; the Veil page carries the current status and timeline.

Response times can vary under load. Sign in for browser chat, or create an Axion API key for CLI and API access.

Web Chat

Chat with Lumen directly in your browser at axion.amplifiedsmp.org/chat — no install needed.

Features

Multiple conversations — sidebar shows all your chats; click to switch, rename with ✎, delete with ✕
Chat history — conversations are saved to your Axion account and sync across signed-in devices. Use the search box above the list to filter by title.
Edit & retry — click ✎ on any message to edit it, or use the retry button to regenerate the last response.
Stop generation — click the stop button (■) to cancel a response mid-stream.
Export — download any conversation as JSON via the ↓ button in the topbar.
Offline-ready — installable as a PWA (look for the install prompt in your browser).

Keyboard shortcuts

Enter — send message
Shift+Enter — new line
Escape — close sidebar / cancel edit

Rate limits

Free account — required for hosted model access; account-wide $0.125 / week and $0.05 / rolling 2h, measured from input and output tokens. Both are lazy-start — the clock only starts on your first request. Sign in at keys to create up to three keys.
Pro account — $7 / month, with account-wide $1.25 / week and $0.50 / rolling 2h included usage plus unlimited active keys.
API credits — redeemed credits cover token usage after the included allowance.

API Keys & Limits

Axion Labs hosts Lumen via an OpenAI-compatible API. Get a key at axion.amplifiedsmp.org/keys.

Tiers

Tier	Limit	How to use
Free account	$0.125 / week · $0.05 / rolling 2h · up to 3 keys	`/login` or `/axion-key axion-sk-…`
Pro account ($7/month)	$1.25 / week · $0.50 / rolling 2h · unlimited keys	Upgrade in Settings

OpenAI-compatible endpoint

        # Base URL

        https://api.amplifiedsmp.org/v1

        # Example request

        curl https://api.amplifiedsmp.org/v1/chat/completions \

          -H "Authorization: Bearer axion-sk-…" \

          -H "Content-Type: application/json" \

          -d '{"model":"lumen","messages":[{"role":"user","content":"Hi"}]}'

Python (openai library)

        from openai import OpenAI

        client = OpenAI(api_key="axion-sk-…", base_url="https://api.amplifiedsmp.org/v1")

        resp = client.chat.completions.create(model="lumen", messages=[{"role":"user","content":"Hi"}])

Teams

Create a team to share API keys with collaborators. All keys are tracked individually so you always know who's using what.

Creating a team

Go to axion.amplifiedsmp.org/keys and sign in
Click + New team in the sidebar
Give the team a name and click Create team

Inviting members

Click Invite member in the topbar when a team is selected. They'll get an email with a link — when they click it and sign in, they're added to the team automatically.

Team API keys

While viewing a team, click + New Key to create a key scoped to that team. Team keys appear under the team, not under personal keys. Any member can create keys; only the team owner can delete the team.

Installation

Axion is a Node.js CLI published on npm. You need Node 18 or later.

        # Install globally

        $ npm install -g @axion-labs-ai/quark-cli

        # Or use the install script (checks Node, retries with sudo on EACCES)

        $ curl -fsSL https://axion.amplifiedsmp.org/install.sh | sh

        # Verify

        $ axion --version

Tip: If axion isn't found after install, make sure your npm global bin directory is on your PATH. Run npm bin -g to find it.

Requirements

Node.js 18+ — required
An account or provider key — use /login for Lumen, or supply an API key for any of the third-party providers below.
Python 3 (optional) — needed for computer use / screen overlay
Tesseract (optional) — for OCR / find-by-text in computer use

Run /doctor inside Axion after install to check your environment.

First Launch

Run axion to start. If no API key is configured, a wizard walks you through setup.

        $ axion

          Welcome to Axion!

          Pick a model to get started:

          › Claude · GPT-4o · Gemini · Groq · Lumen (free)

Select Lumen and sign in with /login, or paste an API key for another provider when prompted.

Sign in to Axion Labs

Run /login inside Axion to open a browser tab, sign in to your Axion Labs account, and automatically save an API key — no copy-pasting needed.

        › /login

          Opening browser… click Authorize in the tab that opens.

          Logged in as you@example.com · API key saved.

Manual setup

        # Set a third-party API key

        › /api claude sk-ant-…

        # Or set your Axion key directly

        › /axion-key axion-sk-…

        # Skip to the free model

        $ axion --model lumen

Models & API Keys

Axion supports multiple providers. Switch between them at any time with /model.

Requires key

Claude

/model claude

Anthropic's Claude models. Get a key at console.anthropic.com.

Requires key

GPT-4o

/model gpt

OpenAI models. Get a key at platform.openai.com.

Requires key

Gemini

/model gemini

Google Gemini. Free tier at aistudio.google.com.

Requires key

Mistral

/model mistral

Mistral's hosted models. Get a key at console.mistral.ai.

Free tier

Groq

/model groq

Fast inference on Llama 3.3 70B. Generous free tier at console.groq.com.

Free · Local

Ollama

/model ollama

Any model running locally via Ollama. No key needed.

Requires key

OpenRouter

/model or

100+ models through one API. Keys at openrouter.ai.

Switching models mid-session

        › /model claude

        › /model lumen

        › /model groq

Conversation history carries over when you switch — the new model gets full context.

Commands

Type / followed by any command. Tab completes the name. Run /help for the full list in-session.

Account & keys

Command	What it does
`/login`	Sign in to Axion Labs via browser — auto-creates and saves an API key
`/axion-key <key>`	Save your Axion API key manually
`/axion-key remove`	Clear your Axion API key
`/api <provider> <key>`	Set a third-party API key (claude, gpt, groq, mistral, gemini, openrouter)
`/model <name>`	Switch model — claude, gpt, lumen, gemini, groq, ollama…
`/models`	List all available models and custom endpoints
`/endpoint <name> <url> [model] [key]`	Add a custom OpenAI-compatible endpoint
`/thinking [on\|off\|<tokens>]`	Toggle extended thinking (Claude only)

Chat & session

Command	What it does
`/mode <name>`	Switch agent mode: ask, plan, bypass
`/adviser [model\|auto\|off]`	Set or disable the adviser model
`/system [text\|clear]`	Set extra system-level instructions
`/include <file>`	Pin a file into every message (no args = list)
`/btw <question>`	Quick side question without breaking flow
`/retry`	Re-run the last message
`/compare [m1,m2,...] <prompt>`	Compare two or more models side by side
`/review`	Code review of the current git diff
`/goal <description>`	Loop until the agent meets a condition
`/compact`	Summarise and compress history to save tokens
`/cost`	Show session token usage and estimated cost
`/pr [context]`	Draft a PR title and body from recent commits
`/clear`	Clear conversation history
`/exit`	Quit

Files & history

Command	What it does
`/run <cmd>`	Run a shell command and feed output to the agent
`/undo`	Restore the last overwritten or deleted file
`/rewind [list\|<n>]`	Undo the last n turns of file changes
`/export <filename>`	Save the current chat as a markdown file
`/save <name>`	Save the current chat session
`/resume <name>`	Resume a saved chat (no args = list all)
`/history <query>`	Search message history
`/copy`	Copy the last AI response to clipboard
`/copy-block <n>`	Copy the Nth code block from the last response

Skills & automation

Command	What it does
`/skills`	List all skills
`/skill <name> <instructions>`	Create a skill from a description
`/skill-delete <name>`	Delete a skill
`/macro record\|stop\|play\|list\|delete`	Record and replay sequences of messages
`/schedule list\|add\|run\|remove`	Manage scheduled AI tasks
`/remember [text]`	Save a persistent note (no args = list all)
`/forget <n>`	Remove a saved note by index

Computer use & vision

Command	What it does
`/computer [on\|off]`	Toggle screen control (alias: /cu)
`/ss [question]`	Screenshot and describe the screen
`/vision <model>`	Set the vision model for computer use
`/voice`	Record voice — transcribed by Whisper
`/img-gen <prompt>`	Generate an image via OpenAI

Integrations

Command	What it does
`/mcp browse`	Browse the MCP marketplace
`/mcp install <id>`	Install an MCP server from the marketplace
`/mcp toggle <name>`	Enable or disable an installed MCP server
`/mcp reload`	Refresh the remote catalog
`/oauth connect\|list\|revoke`	Link GitHub, Google, Notion, or Slack
`/discord token\|start\|stop\|status`	Discord bot mode
`/web [port\|stop]`	Open or stop the web UI
`/plugin …`	Browser, Docker, GitHub, and database plugins

Code vs Chat Mode

Axion has two modes you can switch between at any time.

Code mode (default)

The agent has access to all tools: reading and writing files, running shell commands, searching the web, using the browser, git operations, and more.

$ axion # starts in code mode

Chat mode

No tools — just a conversation. Faster and better for questions, brainstorming, and writing where you don't need the agent touching files.

        $ axion --chat  # start in chat mode

        › /chat  # or switch mid-session

Memory

After each session, Axion extracts key context and injects it into the next session's system prompt automatically.

It remembers things like project decisions, file paths, preferences ("always use TypeScript"), and anything you explicitly tell it to remember.

        › /memory  # view saved memories

        › /memory clear  # clear all

        › /remember always use pnpm instead of npm

Memories are stored locally in ~/.axion/memories.json — they never leave your machine unless you're using a cloud API.

Discord Bot

Run Axion as a Discord bot so your team can queue tasks from any channel.

        # Set your bot token

        › /api discord <bot-token>

        # Start the daemon

        $ axion-discord

        # Override the model

        $ axion-discord --model lumen

Once running, message the bot in Discord. Tasks run in a queue — one at a time. Members can cancel the current task by typing cancel.

Get a bot token from the Discord Developer Portal. Enable "Message Content Intent" under Bot settings.

Chrome Extension

Connects your browser to Axion so the agent can read pages, click elements, and fill forms.

Setup

Start the Axion web UI: /web inside Axion or run axion-serve
Install the extension from its standalone GitHub repository — load unpacked from the repository root
In extension Settings, enter the local URL and the short-lived extension import token printed by /web

What it can do

Read the full text and DOM of any page
Click buttons and links after one-time approval
Fill and submit forms after one-time approval
Take screenshots after one-time approval
Navigate to HTTP(S) URLs after one-time approval