Skip to content

Releases

MegooColab 0.23.0

Choose a tag to compare

@github-actions github-actions released this 16 Nov 11:24

MegooColab 0.23.0 lets you hear new notifications, makes the task drawer as wide as you like, and keeps the selected result in view in Search.

New

  • A chime when a new notification arrives. While MegooColab is open in a browser, a short two-tone chime plays whenever the bell gets a new notification, so you notice it without watching the bell.
    • It plays whether the tab is in front or in the background. With several MegooColab tabs open, only one of them chimes, and notifications that arrive within 2 seconds of a chime don't chime again (the bell still counts them).
    • Browsers allow sound only after you have clicked or typed in a tab, so a tab you just reloaded stays silent until then.
    • Settings → Notifications has a new Sound card: Play sound to hear it, and Turn off or Turn on. The choice is kept per browser, and the sound is on until you turn it off.
  • Drag the task drawer's edge to resize it. On a desktop, the task drawer that opens from the board or the backlog has a grip on its left edge. Drag it to make the drawer wider or narrower, from 720 px up to nearly the whole window.
    • The width is kept in your browser and used the next time a drawer opens, on the board and the backlog alike.
    • Double-click the edge to go back to the usual width. From the keyboard, focus the edge and use the arrow keys, Home and End.
    • On a narrow screen the drawer stays full width, as before.

Improved

  • Search's group headings stay pinned, and the selected result stands out. In the Search popup (Ctrl+K), the heading of each group (Tasks, Documents, …) stays at the top while the list scrolls, and the selected result has a cyan bar and tint.

Fixed

  • The arrow keys keep the selected result in view in Search. Before, pressing ↓ past the bottom of the list selected results you could no longer see. The list now scrolls just enough to show the selected result, and on the last one the "See all results" link too. A new search starts again at the top.
    • A mouse resting over the list no longer takes the selection back while the list scrolls under it; moving the mouse still selects a row.
    • Screen readers now hear which result is selected.

Upgrading

Pull and restart as usual. There's no database change and no new setting. Runner 0.10.0 comes out at the same time and is worth updating to, mostly on Windows (see its note), but MegooColab 0.23.0 works with older runners too.

Runner 0.10.1

Choose a tag to compare

@github-actions github-actions released this 16 Nov 19:53

megoocolab-runner 0.10.1 gives Claude Code agents the MCP servers you set up in the Tool Hub, as OMP and Codex agents already had.

Update

  • Windows (Scoop): stop the service first, then update and start it again:
    megoocolab-runner service stop; scoop update megoocolab-runner; megoocolab-runner service start
  • macOS and Linux: run the install command again: curl -fsSL https://raw.githubusercontent.com/mmhy2003/MegooColab/main/runner/install.sh | sh, then restart the service if it runs as one: megoocolab-runner service stop && megoocolab-runner service start.

Fixed

  • Claude Code agents now see the Tool Hub's MCP servers. Claude Code runs start without your personal Claude settings, so they don't leak into runs, and that also meant they loaded none of the MCP servers in ~/.claude.json, the hub's included. The runner now hands each Claude Code run the hub's servers itself, when a session starts, is resumed or is reloaded.
    • A read-only run gets only the servers marked safe for read-only runs, as with OMP.
    • Servers you added to ~/.claude.json by hand still don't reach runs. Bring one into the Tool Hub with import or Take over to share it with your agents.
    • Codex agents are unchanged: Codex loads the hub's servers from its own config.toml in every run.

Runner 0.10.1 works with any MegooColab server.

Runner 0.10.0

Choose a tag to compare

@github-actions github-actions released this 16 Nov 11:21

megoocolab-runner 0.10.0 keeps the background service running on Windows, restarting the runner after a crash, and says in runner.log why it didn't start when it can't.

Update

  • Windows (Scoop): stop the service first, then update and start it again:
    megoocolab-runner service stop; scoop update megoocolab-runner; megoocolab-runner service start
  • macOS and Linux: run the install command again: curl -fsSL https://raw.githubusercontent.com/mmhy2003/MegooColab/main/runner/install.sh | sh, then restart the service if it runs as one: megoocolab-runner service stop && megoocolab-runner service start.
  • macOS: run megoocolab-runner service install once after updating, so the service gets the new restart rules below.

New

  • Windows restarts the runner after a crash. At sign-in, the startup entry now starts a hidden supervisor that starts the runner and starts it again if it crashes: after 5 seconds at first, then waiting longer each time, up to 2 minutes. It doesn't restart a runner you stopped, or one that stopped because it needs login again.
    • Task Manager therefore shows two megoocolab-runner.exe processes. service stop stops both.
    • The runner still starts only after you sign in: a computer waiting at the sign-in screen (for example after Windows Update) has no runner.

Improved

  • service install does the whole job. It refuses until login has been done, then registers the service, starts the runner and shows service status. Running it again replaces the installation. service start does nothing when the runner is already running, so the commands shown in MegooColab still work.
  • runner.log says what happened. Every line starts with the time. Each start writes starting megoocolab-runner <version> (pid N), and a failure to start writes error: …, for example error: not logged in …. On Windows, the supervisor's restarts and any crash trace go to the same file. If runner.log can't be opened, the runner writes to megoocolab-runner-start.log in the temp folder.
  • service status on Windows warns when autostart is broken: it shows the stored startup command, and says so when that file no longer exists or the entry was disabled in Task Manager → Startup apps. service install turns a disabled entry back on.

Fixed

  • The runner no longer stays in a console window at sign-in on Windows, for example when Windows Terminal is the default console. The startup entry now always moves the runner to a hidden process; a brief console flash is still possible.
  • A missing or broken runner.json no longer makes the service restart forever. The runner exits with status 78, as for a revoked token, and the service waits for login.
  • macOS no longer restarts the runner in a loop when it needs login. launchd still restarts it after a crash, but a runner whose token was revoked or replaced, or that isn't logged in, now stays down until login and service start.
  • On Linux, service install replaces a unit that systemd had marked as failed.

Runner 0.10.0 works with any MegooColab server.

MegooColab 0.22.1

Choose a tag to compare

@github-actions github-actions released this 14 Nov 14:47

MegooColab 0.22.1 fixes dragging a card into another column on the board when the columns are long.

Fixed

  • A card dropped into another column lands there, wherever you let go of it. Before, a card dragged out of a long column only stayed in the new column if you let go near that column's middle; anywhere else it went back to where it came from. On a long board, that middle could be far below the screen.
    • The card now goes into the column under the pointer: at the card you let go over, next to the nearer card between two, or at the end below the column's last card or in an empty column. The column's header takes a card too.
    • Moving cards with the keyboard works as before.

Upgrading

Pull and restart as usual. There's no database change and no new setting, and the runner doesn't need updating.

MegooColab 0.22.0

Choose a tag to compare

@github-actions github-actions released this 14 Nov 13:25

MegooColab 0.22.0 tells you when a task is done: your own and your agents' tasks, and, if you like, every task in your team's projects.

New

  • A notification when a task reaches Done. When a task moves into a Done column, MegooColab now tells the people it concerns, so you no longer keep the board open to see what your agents have finished.
    • Your own tasks: a task assigned to you, or to one of your agents, tells you "PAY-42 is done", with its title and who finished it.
    • Your team's tasks: everyone else on the project's teams hears about it too.
    • Settings → Notifications has a row for each: A task assigned to me or my agent is done (on in the app and by email) and A task in my team's projects is done (on in the app only, since a busy team finishes many tasks a day). Each can go to the app, email, push and Telegram, like the other rows.
    • The notification says who moved the task: a person, your agent (for example "moved to Done by laptop-claude (your agent)", when a chat's agent finished it), or "when its workflow finished".
    • You're never told about a task you moved to Done yourself, and you get one notification even when both rows apply to you. Moving a task between two Done columns tells nobody; a task that is reopened and finished again tells you again.
    • A chat's own rounds don't notify: the chat is where you follow those.

Upgrading

Pull and restart as usual. There's no database change and no new setting, and the runner doesn't need updating. The new notifications start with the defaults above; change them on Settings → Notifications.

MegooColab 0.21.0

Choose a tag to compare

@github-actions github-actions released this 14 Nov 10:45

MegooColab 0.21.0 lets agents use the files people put in a task's Attachments section, such as a logo to add to the project.

New

  • Agents can get the files in a task's Attachments. Until now a file added with "+ Attach a file" never reached the agent: it wasn't in its instructions and no tool could fetch it, so an agent asked to use an attached logo said it couldn't see one.
    • A run on a task with attached files is told they are there: name, type, size and id of each (up to 20; a task with more says how many more).
    • Two new tools fetch them. list_attachments lists a task's files; get_attachment gives one file's download link, which the agent saves wherever it needs the file (for example into the project's assets). An image is also shown to the agent as a picture, so it knows what it is using.
    • get_task lists the task's files too, so an agent sees files added after its run started, and your own agent connected over MCP gets the same tools.
    • A download link works for 15 minutes, for that one file only, without signing in. It can't be changed to reach another file, and it stops working if the file is deleted.
    • Images sent with a message to the agent work as before: they go with the run the message was for, and aren't offered by these tools.

Upgrading

Pull and restart as usual. There's no database change and no new setting, and the runner doesn't need updating.

The download links use your server's PUBLIC_URL, the same address agents already reach MegooColab's tools on, so no network change is needed.

MegooColab 0.20.0

Choose a tag to compare

@github-actions github-actions released this 13 Nov 16:37

MegooColab 0.20.0 lets each workflow stage run on its own model and effort, so an agent can design and plan on one model and build on another.

New

  • A model and effort for each workflow stage. An agent's settings (Agents → the agent) have a new Model by stage table under its Model and Effort, with a row each for Design, Plan, Build (steps and their fixes), Review (step, final and pull request reviews) and Finish. For example, set Design and Plan to Fable 5.2 and leave the rest "As above" to build on Opus.
    • Each row offers the models the agent's own program lists, with the effort levels each model has, or a typed name. A row that sets a model but no effort uses that model's own default effort. Rows with their own choice are outlined in violet.
    • Every run of a stage starts on its choice, also when it continues an earlier session: a session designed on Fable goes back to the agent's own model for building.
    • A task's Activity already says which model and effort each run used, so you can see the stages switch.
    • Chats, scheduled jobs, answers to other projects and Knowledge Base drafts keep using the agent's own Model and Effort.
    • Changing the agent's own model resets a stage that set only an effort the new model doesn't have, and says so. The same goes for a stage's own model and its effort.

Upgrading

Pull and restart as usual. The update adds a database column for the stage settings and one for what each run was started on; it is applied automatically when the server starts, and there's no new setting.

Update the runner on each machine to 0.9.0 to use stages with their own model. Until a machine's runner is updated, runs of a stage with its own model wait, and the agent shows "runner too old to choose a model per stage — update megoocolab-runner". Agents without stage settings, and stages left "As above", work with the older runner as before.

MegooColab 0.19.0

Choose a tag to compare

@github-actions github-actions released this 13 Nov 13:15

MegooColab 0.19.0 makes steering an agent quicker, has agents check current library documentation, gives agent replies the full width on large screens too, and puts MegooColab's icon on a mockup's tab.

New

  • The comment box sends to the task's agent by default. While a task isn't in a Done column, and you may message its agent, the "Send to " switch starts on, and after each message it is on again. You no longer click it before every instruction, and an instruction can't go out by mistake as a comment the agent never reads.
    • To write a comment just for people, turn the switch off for that message: after you send it, the switch is back on. Mentioning the agent with @ still sends to it.
    • In a Done column the box starts as an ordinary comment, as before. So does it for people who can't message the agent.
  • Agents check library documentation before relying on memory. Every agent session (task steps, reviews, freestyle chats, scheduled jobs, answers to other projects and Knowledge Base drafts) is now told: when you need facts about a library, framework, SDK, API or command-line tool, look in the Knowledge Base first, then use Context7 if you have it, even for well-known libraries. A session without Context7 carries on as before. The instruction is built in, so projects with their own edited instructions get it too.

Improved

  • Agent replies use the full width on every screen. In 0.18.3 a reply's time moved next to the agent's name on a phone. Tablets and laptops now get the same layout: the time sits after the name or at the end of an event line, and replies, run summaries, fix-round summaries and pull request excerpts reach across the whole Activity column. In an open task on a laptop, a reply gets about a third more width.

Fixed

  • A mockup opened full screen shows MegooColab's icon on its tab. The tab used to show no icon, because the mockup page is locked down and the browser couldn't load the site's icon for it. A mockup that sets its own icon keeps it.

Upgrading

Pull and restart as usual. There's no database change and no new setting, and the runner doesn't need updating.

To let agents use Context7: add Context7 to your Tool Hub (if it isn't there yet), turn on Safe in read-only runs for it, and push it to your machines. Brainstorming, planning and reviews run read-only, so without that setting their Context7 calls are refused and they carry on without it.

Runner 0.9.0

Choose a tag to compare

@github-actions github-actions released this 13 Nov 16:32

megoocolab-runner 0.9.0 starts each workflow run on the model and effort its stage asks for, even when the run continues an earlier session.

Update

  • Windows (Scoop): stop the service first, then update and start it again:
    megoocolab-runner service stop; scoop update megoocolab-runner; megoocolab-runner service start
  • macOS and Linux: run the install command again: curl -fsSL https://raw.githubusercontent.com/mmhy2003/MegooColab/main/runner/install.sh | sh, then restart the service if it runs as one: megoocolab-runner service stop && megoocolab-runner service start.

New

  • A run can carry its own model and effort. With MegooColab 0.20.0 and later, a stage that has a model of its own (Agents → the agent → Model by stage) sends it with each run, and the runner puts the agent's session on that model and effort.
    • Continued sessions switch too. A session an earlier stage put on another model goes back to the agent's own when the next stage asks for it. To know what "the program's own model" is, the runner notes the model each program starts a new session on and keeps it in its cache, so it survives a restart.
    • Until it knows, a run that has to go back to the program's own model starts a new session instead of continuing the old one, with everything the continued session would have been told. This can happen once per program after updating.
    • A stage with a model but no effort runs on that model's own default effort.
    • Programs that offer no model list keep continuing their sessions as before.

The runner works with any MegooColab server: an older server never sends a run its own model, and runs use the agent's model as before.

MegooColab 0.18.3

Choose a tag to compare

@github-actions github-actions released this 12 Nov 15:19

MegooColab 0.18.3 stops build steps from getting stuck over files git ignores, makes agent replies easier to read on a phone, and fixes the count of new entries on the ↓ button.

Fixed

  • Ignored files no longer send finished work back to be committed.
    • What went wrong: since 0.18.1, a step whose agent had committed its work could still be sent back with "left its changes uncommitted" if the run created or removed files that git ignores, such as __pycache__ folders. The agent found nothing to commit, cleaned those folders up, and the task stopped with "The agent left its changes uncommitted, so they can't be reviewed."
    • Now: only changes that git status actually shows as uncommitted send the work back to its agent. Ignored files and MegooColab's own .megoocolab/ folder don't count.
    • Needs runner 0.8.2, which reports what is left uncommitted. With an older runner, work goes straight to its review, as it did before 0.18.1.
  • Agent replies use the full width on a phone. In tasks and chats, the time of an agent's reply used to take its own column on the right and squeezed the reply into half the screen. On a phone the time now sits next to the agent's name, and the reply's box starts under the row's dot. Event lines show their time after the text, and run summaries, fix-round summaries and pull request excerpts also use the full width. Tablets and larger screens look as before.
  • The ↓ button counts past 50. When you scroll up in a busy task, the number of new entries on the ↓ button stopped at 50. It now shows the real number up to 99, then "99+". Chats show "99+" past 99 too.

Upgrading

Pull and restart as usual. There's no database change and no new setting. Update the runner on each machine to 0.8.2 so that work left uncommitted is caught without tripping over ignored files. A task already waiting with "The agent left its changes uncommitted" continues when you press Retry.