Recent events across all runs, media, memories and bridge jobs.
Loading…
Search
Full-text search across run messages and agent memories.
Mission templates
Reusable task blueprints. Use {{variable}} placeholders in title or prompt.
+ New template
Loading…
Fill in variables
Connections
Provider credentials live in an encrypted vault inside your worker (AES-256-GCM, keyed by your MASTER_KEY secret) — pasted once here, never in config files. Since early 2026, Anthropic and OpenAI only permit third-party apps to authenticate with API keys; consumer-subscription sign-in is restricted to their own apps. Google is the exception: sign in with your Google account and calls route through Vertex AI on your GCP project.
Operator alerts
Get notified on run errors, budget warnings and video completions via ntfy or Telegram.
Ask it anything from your phone. It answers from live data and reads the vault. Anything that would change something is parked for your approval here — the chat can answer, not act.
checking…
Account
A bearer token for scripting or the CLI bridge. Rotating replaces the old one immediately — copy it now, it is shown once.
Users
Add teammates and set their role. Members use missions; admins manage everything. A temporary password is shown once — share it, they change it in Account.
＋ Add user
Fleet analytics — 30 days
Token spend is computed from each agent's configured rates. Edit rates in an agent's Config tab to match current provider pricing.
Loading…
Files
R2-backed · organize into projects & folders · available to agent briefings · max 25 MB each
Calendar
Tasks that ran · files created · scheduled work — filter by department, click a day to act
● ran ● scheduled ● files
Media lab
OpenAI images · Google Veo video · results stored in R2
Generate image
Generate video
Library
Add an agent / LLM
New agent will appear in the neural viz. The orchestrator routes each subtask to the agent whose strengths best match.
Pick from the live catalog or choose "Custom…" to type any model string.
Knowledge Base
Facts here are injected into every agent's context — the crew treats them as ground truth about ALMA's products.
+ Add knowledge
Loading…
Missions
Live and recent agent runs across the entire fleet.
Ongoing
Loading…
Recent
Loading…
Departments & Projects
Organize agents and people into departments and projects. Missions can be scoped to a project so only its designated agents work on them.
Loading…
Voice settings
OpenAI Realtime — requires an OpenAI API key in Connections.
Say this to ALMA to unlock full authority on a shared device. When it's not you, ALMA still answers and runs read-only work, but anything that changes things, sends messages, or spends waits for your approval.
🧠 Brain Vault
Loading…
Welcome
scroll / pinch to zoom · drag to pan · click a node to open
Backlinks
—
Outgoing links
—
Memory fabric
…
◉ Owner profile
A note about you — your priorities, the metrics that matter, and how you like things communicated. It's woven into every agent's prompt and the daily briefing, so the brain grows more "you". Agents also propose additions after missions (you approve them in Learning review).
Keep it under ~4000 characters — all of it is fed to every agent, so shorter and sharper works best.
🗓 Standing missions
Recurring missions the brain runs on its own. Edit the prompt, change the cadence, pause without losing the schedule, or have each result delivered to you.
Loading…
🧠 Learning review
After a mission, agents propose durable memories and reusable skills. Approving is the promotion gate — approved memories join the memory fabric; approved skills load into every agent.
MISSION CONTROL