Changelog

Command Center keeps itself current. Releases arrive as cryptographically signed auto-updates in the background — you never download an update by hand. How updates work explains the mechanics; this page is the record of what ships.

Releases are listed newest first. Later entries group changes as Added, Changed, and Fixed; 0.1.0, as the initial release, lists the launch surface as a whole.

Prefer to be told? The release feed carries every entry on this page in any feed reader.

0.1.0 — Initial release

The first public release. Command Center installs as a Windows desktop application, provisions its services and its integrated graph database locally during first-run setup, and runs on your own Claude subscription. Everything below is live in 0.1.0 — Pro covers the platform and the intelligence desk; Ultra adds the production modules and Voice.

• Mission control — a live mission board where missions decompose into operations and are staffed by a commander-led fleet of agent sessions. Success conditions gate when a mission counts as done; one click pauses the whole fleet, and every control action lands in an audit log on your disk.
• Date-addressed scheduling — a calendar whose entries dispatch an agent at a moment in time, in the background or into a terminal workspace you can watch, on an autonomy level you set per entry. Describe a week in prose and an agent turns it into entries; nothing an agent schedules fires without the dispatch switch you hold.
• Email — a full IMAP/SMTP cockpit with a triage queue, an agent that researches and drafts replies, and approval on every outbound message.
• The terminal fleet — persistent Claude Code sessions in a split-grid terminal workspace, each grounded with curated context. Sessions survive reloads, pop out into their own windows, and resume after a restart.
• The intelligence desk — harvest RSS/Atom feeds, public X accounts, YouTube channels, and newsletter inboxes with no API keys, deduplicated and filed locally. Finished formats in your voice: briefings, breaking items, deep reports, roundups, and newsletters. Anything outbound stays a draft until you release it.
• Video intelligence — paste a YouTube URL, or schedule a whole channel, and receive an 8-section deep analysis with frame references timestamped to the second.
• Cortex — the knowledge graph: folders, documents, and extracted relationships as a browsable map, at three depths (structure, entities, full). The default path costs zero external model tokens; semantic vector search is an opt-in with your own key.
• Memory — six separated memory layers with hybrid keyword-plus-meaning search. Entries can be updated, superseded with lineage, or deleted, and every store is visible in the app.
• Agents, skills, and routines — a roster of pre-built specialist agents, procedural skills, and scheduled routines, each a readable Markdown file you can edit or clone. Changes go live on the next run, without a restart.
• Voice (Ultra) — a summonable voice assistant that drives the platform hands-free: it opens views, arranges windows, drives terminal sessions, and walks the knowledge graph, on a tool reach you curate. The conversational AI runs on your own provider account.
• Dictation (Pro) — system-wide speech to text on a local model: hold a hotkey in any application, speak, and the text lands at the cursor. Fully offline, no account, no per-word cost.
• AI Video Generation (Ultra) — two complete idea-to-video pipelines plus raw-footage prep, operated from the cockpit.
• Atlas Engine (Ultra) — the knowledge-product factory: an idea plus raw material becomes a finished two-component knowledge product.
• Updates and entitlement — signed auto-updates over an authenticated channel, a single sign-in, and offline tolerance: a network blip never locks you out.