What's new

Changelog

Notes for every released build. Auto-updates ship the same content quietly in the app.

  1. v0.3.0

    In 0.2.0 one agent at a time did things. In 0.3.0 several agents can work together on one request, agents can work on your own folders, and everything you add to Astros installs one way, as a plugin, and runs sandboxed.

    Several agents in one conversation

    Assign the work — mention two or more agents anywhere in a message and press Enter. One agent takes the lead and splits the request into tasks, the others take their share, and a plan card under your message shows each task, who owns it, its progress and the running cost. Type while it runs to add a steering note; Stop ends it. When the job is only one task, you simply get a reply.

    Ask each instead — ⌥Enter sends the message to every agent it mentions, each answering separately. From the search bar, @a @b … offers Plan with A + B and Ask A + B separately.

    Agent profiles — click an agent's avatar or name in a reply to see who it is, how it is set up and what it has done in this conversation, with a link to its settings.

    Agents that work in your folders

    Working directories — give an agent folders on its Preferences page, each Read or Read & write. A conversation starts with its agent's folders as chips above the message box; add, remove or change them there. Another agent in the conversation asks before it touches one (Allow read, Allow read & write or Deny, for this conversation only).

    Edits you can undo — a file an agent changed shows in the reply with its +/− counts. Open shows the diff, Revert puts the file back, and Undo revert brings the change back again. Claude Code and Codex agents start in the first folder on the list.

    Plugins

    One way to install — + Install… in Preferences installs a plugin from a folder, a URL, a catalog, npm or GitHub. A plugin can bring agents, models, tools, MCP servers and skills; before it installs, Astros shows what it can reach and which accounts it will ask for, and an update asks again only when it wants more. A skill package from npm or GitHub installs the same way. Plugins in Preferences lists everything installed, built-ins included.

    Sandboxed — installed plugins run isolated and can reach only the sites they declared. On macOS and Linux, local MCP servers run sandboxed too, limited to the folders and sites you approve in the add dialog; servers launched with npx, uvx and similar are pinned to the version you approved. On Windows a one-time Enable isolation step turns this on, and local MCP servers wait until it is done.

    Tools you can call by name — ElevenLabs is now a set of tools (speak, dialogue, sound effects, music, transcribe, change voice, isolate voice) any agent can be given, and @elevenlabs.say Good morning runs one straight from the search bar. Luma, Kling, Stability and Leonardo arrive the same way, and @media uses whichever of them you grant it.

    In the conversation window

    • Queue a follow-up — press Enter while a reply is still running and your message waits its turn. Edit, Send or remove it from the strip above the message box.
    • Drafts stay put — a half-written message stays with its conversation when you switch, and typing anywhere in the window lands in the message box.
    • Agents can ask you — an agent can stop mid-reply to ask a multiple-choice question.
    • Longer runs — an agent can take up to 90 tool rounds per reply (adjustable in Preferences), and a reply that hits the limit gets a Continue button that picks up where it stopped.
    • Files — ⌥⌘B opens a conversation's files; right-click one to reveal, open or copy its path; ⤢ shows a preview full screen. Click an image, or ⤢ on a video, for a full-window viewer with zoom and ←/→.
    • Images and PDFs — you can now attach them for GPT and other OpenAI-compatible models too, and tools can hand images and PDFs back to the model.

    In the search bar

    • Browse folders by keyboard — Tab on a folder in @files puts you straight in its tree; Tab or → on a file opens it beside the tree.
    • @files operators — ^src matches a folder or file name starting with src, and !test excludes matches.
    • Parquet files preview as tables.

    Models and web search

    @claude.haiku runs Claude Haiku 5.5, @gemini.gemini-4-argon adds Gemini 4 Argon, DeepSeek gains a reasoning-effort setting, and ElevenLabs speech moves to Eleven v4. Web search is now a tool you grant to the agents that should search; Claude and Gemini use their own built-in search, and a reply shows each search where it happened. With no default chat model set, Astros asks you to choose one instead of quietly using Gemini.

    Also

    • Preferences is reorganized: a Your Astros group lists Agents, LLMs, Tools, Servers, Skills, Harnesses, Mentionables and Plugins, each with its own list.
    • Write a skill without leaving Preferences: Skills › + Add skill › Write manually….
    • A file you attach that macOS won't let Astros read now says why, and where to allow it.
    • Astros's local database moves to PostgreSQL 18; existing data migrates on first launch.
    • Plugin authors: plugins now declare their parts in manifest.json and export them with defineMentionable / defineTool; the old defineAgent is gone.

    Multi-agent conversations and installing plugins are Pro. Sync is now its own subscription, separate from Pro; your first sign-in starts a 30-day trial of both.

    Upgrading on Windows: your local MCP servers won't start until you run the one-time Enable isolation step that Astros offers.

    macOS (Apple silicon and Intel) and Windows. If you run 0.2.0, the in-app updater will offer this build.

  2. v0.2.0 — Agents that do things

    In 0.1.0, the agents behind the bar answered. In 0.2.0 they do things: call tools, load skills, hand work to each other, write files, and keep going after you look away.

    Agents that use tools

    MCP servers — add any MCP server in Preferences: paste its config and Astros fills in the form. Local and remote (HTTP) servers both work, with OAuth sign-in where a server asks for it. You approve a server before it first runs, choose which agents may use it, and can talk to one directly with @mcp.<name>.

    Skills — Astros reads the SKILL.md skills already on your machine (~/.claude/skills, ~/.agents/skills, and its own ~/.config/astros/skills). An agent loads one when the task calls for it, or you invoke it yourself by typing /name, in the search bar or in a conversation's message box.

    Built-in tools — three tool servers ship in the box: DuckDB for querying data files, Python for analysis and charts, and a media server for image, audio and video work. @tabular and @media run on them, and any agent you grant them to can use them too.

    Agents you compose

    The Composer — build an agent without code. Give it a name, instructions, the tools and skills it may use, the agents it may search and the agents it may delegate to; it then lives at @<name> like any other. Clone agent… starts from one you already have.

    Delegation — a composed agent can hand part of a task to another agent. The hand-off appears in the reply as a row you can open to read the whole delegated conversation, then step back.

    Runs on Claude Code or Codex — a composed agent can run on the Claude Code or Codex you already have installed and signed in to, or on any ACP agent you add. From there it can still search your Astros agents and delegate to them.

    Create with AI — + New agent ▾ gathers every way to make an agent. Describe one and @astros plans it first, then either writes the code for you to review or pre-fills the Composer. (@astros.create is now the /create-agent skill on @astros.)

    The conversation window

    Conversations have a window of their own. Enter on a conversation row opens it there; Tab still previews it in the launcher.

    • History — recent conversations sit in a panel on the left. Click to switch, ⌘N for a new one, right-click or F2 to rename. A conversation reopens where you stopped reading.
    • Replies keep running — switch away, close the window or dismiss the launcher and the reply finishes in the background, with a notification when it is done.
    • Files — every conversation has a folder. Agents write into it, the files show up in the reply as rows you can open, and a side panel previews HTML, Markdown, images and CSV.
    • Easier to follow — avatars beside each message, tool calls shown where they happened, and a "Working…" line while a reply is still coming.

    In the search bar

    • Folders in @files — folders are results now. Enter opens one in Finder or File Explorer; Tab browses it as a tree without leaving the bar.
    • Profiles — keep separate sets of accounts, such as work and personal. Switch from the chip on the bar or with /profile, and each agent uses the account that belongs to the active profile.
    • Agent commands from anywhere — /tasks:add buy milk runs an agent's command without scoping to the agent first.
    • Smaller things — @emoji browses the whole catalog by category, @tasks separates Active from Completed, and @def now looks words up in Wiktionary.

    Models

    @claude moves to Opus 5.5 and Sonnet 5.5, @openai to GPT-6, @gemini to Gemini 3.8 Flash, and @deepseek to its current models. One reasoning effort setting maps onto each provider's own levels. An agent can now take up to 32 tool rounds in a reply, and the cost shown for a reply counts thinking and cached tokens correctly.

    Also

    • The conversation and Preferences windows remember their size and position.
    • Astros now runs on Electron 44.

    Composed agents, MCP servers and skills are Pro features.

    macOS (Apple silicon and Intel) and Windows. If you run 0.1.0, the in-app updater will offer this build.

  3. v0.1.0 — Astros

    Astros is a launcher — a frameless search bar you summon with ⌘⇧Space, type into, and pass through on your way somewhere else. What makes it different is that everything behind the bar is an agent: a small, addressable thing you can scope to with @.

    Type anything for ranked results across every agent at once. Type @files report to scope to one. = for the calculator, / for app commands. Enter hands off — opens the file, the app, the URL — and Astros gets out of the way.

    What's in the box

    Your machine — @files path-aware fuzzy search over your home directory, @apps for applications, @clip for clipboard history, @snip for saved text snippets, @emoji as a navigable grid, @calc, @def.

    Your browser — @browser searches bookmarks, history and live tabs as one blended set across Chrome and Safari (Edge on Windows). Opening a live tab focuses its window and un-minimizes it.

    The web — @google, @bing, @duckduckgo, @wikipedia, @youtube, and @websearch.

    Your work — @github (issues, PRs, code), @notion, @tencent docs, @docusign, @tabular for spreadsheets and data files.

    AI — @claude, @openai, @gemini, @deepseek, @openrouter, answering in a live preview pane. @media generates and edits images, video and audio; @astros researches across your files, apps and the web.

    Your own stuff — @go for go links, @tasks for a todo list, @history, and @dashboard, which opens a set of pages as a scrolling grid of live previews.

    Beyond search

    • Preview — results open inline: web pages, files, AI conversations. ⌘U toggles rendered and raw.
    • ⌘K action menu on any row — reveal in Finder, copy, rename, attach to an app.
    • Hotkeys — a global summon chord, plus per-agent chords that drop you straight into one agent.
    • Frequently used — a frecency-ranked section on the empty bar.

    Making it yours

    Agents are installable packages, and you can build one by describing it: @astros.create generates it, shows you the actual code to review, and installs what you approved. Agents that need an account can sign in with OAuth rather than asking for a pasted API key — your refresh token never reaches agent code. Any dotted path (@browser.tabs, @openrouter.qwen) is a real place with its own settings page, colour, chords and accounts, so you can pin a model released this morning without waiting for an app update.

    macOS (Apple silicon and Intel) and Windows.


    About the version number. Astros previously reached 1.8.0. The line restarts at 0.1.0 so that 1.0.0 can mark leaving public beta, rather than arriving mid-sequence. This is the same product, not a rollback.

    If you already run 1.8.0, the in-app updater will not offer you this build — it refuses to move to a lower version. Download from astros.cx/download and install over your copy; settings and data stay put, and updates resume normally from here.