The Alfred Black dashboard is where you see everything Alfred is doing for you — your vault, your specialists, your devices, and your services. This guide walks through each section.
The Home page opens with the Vault Nebula — a CSS cloud visualization where each cloud represents a cluster of related vault records discovered by the Surveyor. Clouds are sized by record count and colored by cluster theme, giving you an ambient, data-driven map of your world at a glance.
On your very first login, before the Nebula has data, Alfred displays a First Brief — a short welcome message that explains what’s happening behind the scenes (Gmail backfill, vault initialization) and what to expect in the next few hours as your vault populates.
Below the stat cards, activity charts show recent trends for the Curator, Janitor, and Distiller — giving you visibility into how actively your specialists are working.
An interactive visualization of your vault’s structure. It shows record types as nodes, sized by count, with specialist nodes showing their operational status. Toggle the graph on or off with the switch in the card header.The graph refreshes every 15 seconds when visible.
The Matters page lists your active ongoing concerns — threads of related work Alfred is tracking on your behalf. Each matter shows its linked errands, people, and decisions. You can drill into any matter to see the full timeline.
The Triage page surfaces items Alfred couldn’t confidently classify. Each item shows a summary and suggested routing. Review, reclassify, or dismiss items here. Routing a triage item teaches Alfred for next time via the intuition system.
For each specialist, you can pick a model from the dropdown grouped by provider:
Provider
Key required
Example models
Anthropic
ANTHROPIC_API_KEY
Claude Sonnet 4.5, Claude Haiku 4.5
OpenAI
OPENAI_API_KEY
GPT-4.1, GPT-4.1 Mini, o3, o4-mini
Google
GOOGLE_API_KEY
Gemini 2.5 Pro, Gemini 2.5 Flash
xAI
XAI_API_KEY
Grok 3, Grok 3 Fast, Grok 3 Mini
Meta
via OpenRouter
Llama 4 Maverick, Llama 4 Scout
DeepSeek
via OpenRouter
DeepSeek R1, DeepSeek V3
Mistral
via OpenRouter
Mistral Large, Mistral Small
If you select a model and the required API key is not configured, a dialog prompts you to enter it. You can also enter a custom model ID if your preferred model is not listed.
Models from Meta, DeepSeek, and Mistral are accessed through OpenRouter. Set an OPENROUTER_API_KEY in Credentials to use them. An OpenRouter key also works as a fallback for Anthropic, OpenAI, Google, and xAI models.
The Devices section (accessible from the Home page Quick Actions area) manages OpenClaw device pairing. When a device (phone, tablet, or other client) requests to pair with your Alfred:
The request appears as pending in the Devices section
You can approve or reject it
Approved devices appear in the paired devices list
Four metrics across the top: events processed today, total observations, active instincts, and auto-route rate (the percentage of inputs Alfred handles autonomously).
Inputs that need your routing decision. Each item shows a brief summary, how long ago it arrived, and the best instinct match (if any) with its confidence score. Use the Route dropdown to route the input to a vault destination — this creates an observation and teaches Alfred for next time.
Your active and developing instincts. Each instinct shows its observation count, discretion level, and how many inputs it has auto-routed. Instincts marked as “developing” need more observations before Alfred will use them for autonomous routing.
A chronological log of recent intuition activity: events processed, inputs auto-routed, and reflection outcomes.For more details on how intuition works, see the Intuition guide.
The Credentials page manages API keys for LLM providers used by your specialists. Each credential shows:
Status — Whether the key is set (green check) or missing (gray X)
Description — What the key is used for
Masked value — A preview of the configured key
Used by — Which specialists or services depend on this key
You can set, update, or remove credentials at any time. Changes restart the Alfred worker automatically so the new keys take effect.For more details, see the API Keys guide (for Alfred API keys) and the Your Specialists guide (for provider credentials).
The Logs page streams real-time logs from all services running on your Alfred. Use this for monitoring specialist activity or verifying that everything is working correctly.You can also access logs via the API. See the Monitoring guide.
Links to the Polar customer portal where you can manage your subscription, update payment methods, view invoices, or cancel. See the Billing & Subscriptions guide for the full lifecycle.
Create and manage API keys for accessing the Alfred API. Each key provides full access to your API endpoints. Keys use the alf_ prefix format and are shown only once at creation time.See the API Keys guide for usage examples and best practices.
Navigate to Settings → Terminal to open a web-based terminal with a proper PTY. This gives you full TUI support — you can run interactive CLI tools like htop, vim, or the OpenClaw CLI directly in the browser. The terminal connects to your Alfred instance over a secure WebSocket.
Settings → Terminal now includes a “Download SSH Key” button. This downloads a .pem private key file that lets you SSH directly into your tenant for full CLI access.To connect after downloading:macOS / Linux:
SSH access is useful for running interactive commands like openclaw configure that require a real terminal session. Your SSH key is generated per-tenant during provisioning and stored encrypted on the platform.
Go to Settings → Credentials and add ANTHROPIC_API_KEY with your Claude setup token. This enables Claude models (Sonnet 4.6, Haiku 4.5) powered by your Claude Pro or Team subscription — instead of pay-per-token API credits.To generate your setup token, run claude setup-token on your local machine. The token links your Claude subscription to your Alfred instance.
This is separate from a standard Anthropic API key. The setup token uses your existing Claude Pro/Team subscription, so you are not billed per token.
