Automations (beta)
A job you do again and again, as one board on your canvas: each step a set of chats, you keep the outputs you like, and a conductor chat that runs and changes it for you.
Automations are part of Pro. An automation is a process you run again and again, like "write five reel scripts, I pick the good ones, each one I pick becomes a video". It sits on your canvas as one board: a list per step, and a last Ready list for what came all the way through. Each row is one chat's output. You open it, keep it, drop it, or tell its chat what to change. What you keep moves on to the next step.
Beside the board sits its conductor, a chat that runs the automation for you. Tell it "start one about coffee", "keep 3 and 5", "tell 4 to make it shorter" or "add a step that writes a caption", and it does it.
The conductor is your own Morphy, with running this automation as its main job. It knows what Morphy knows about you, so you can also brainstorm with it, ask it anything, or have it put notes and cards on the canvas. Things it learns about the automation itself it keeps in the automation's folder; only what's about you goes into Morphy's instructions, which every Morphy chat reads.
Automations are in beta. They work, but they're new: expect rough edges, and tell us what you run into in the Discord.
Make one
- Start from a template: open Settings → Community → Automations, pick one, and click Add to canvas. Its page says what it needs on your computer (an AI, a key, a program) and how to get each one.
- Start from scratch: right-click the canvas and pick New automation, or right-click a project in the sidebar, or press Ctrl + K and type "New automation". Right-click inside a project's frame and the automation belongs to that project.
- Or ask Morphy: "make me an automation that turns a topic into three LinkedIn posts, and I pick one". Morphy looks in the templates first.
A new automation runs on an AI you have connected (Claude first, else Codex, and so on). Its conductor opens with a few questions: what it should make, what each run starts from, and where you want to step in. A template's conductor asks what makes it yours instead. Answer what you like and it builds the steps for you.
On the one canvas an automation sits to the right of its project's frame, never inside it. The sidebar lists it under its project, with how many outputs wait for you.
Open it
Click an automation in the sidebar (or find it with Ctrl + K) and it opens in focus with its own tab: the board takes most of the screen, with its conductor beside it. It works like any focus: drag the gap between them to resize, drag a title to swap or split, and drop more chats in. A row's chat you open from the board joins it too. It remembers how you left it.
Esc goes back to the canvas. To see the board on the canvas instead, right-click the automation in the sidebar and choose Show on canvas.
Run it
Click ▶ Run on the board. If the automation needs something to start from (a topic, a file, a folder), it asks first. Above the start button it says what the run will start: how many chats per step, on which AI, and whose plan they use.
The first step's chats start at once, side by side. Each row shows what its chat is doing ("creating script.md · 0:42"), and turns into a title and a line about it when the chat is done.
When outputs wait for you
The board's name row says Needs you · 3, and Next takes you to the next one waiting. On each row:
| What it does | |
|---|---|
| Open (the eye) | Shows its files in Morphy: video, pictures, sound, Markdown and text. Other files open in their own app. A file that runs when opened (a program, a script, a shortcut) asks first: Cancel, Show in folder or Run. |
| Keep (the step's own word, like Make video) | Passes it on. The next step starts on it. |
| Reply | Tells its own chat what to change. It makes a new version (v2, v3…), and the old one stays. |
| Drop | Hides it under "dropped". Bring it back undoes that. |
| ⋯ | Each file, its folder, Open its chat (it becomes a normal chat card), Stop. |
Click a row first to use the keys: N next one waiting, arrows to move, O or Enter open, K keep, R reply, X drop. In the viewer, ← → step through the files, S shows what it was made from, and Esc closes it.
When a run ends
A run ends when only approving is left. To end it sooner, open its run menu and pick Finish — drop what's still waiting. Rows that are already working carry on.
When you approve an output of the last step, Morphy copies it to your own folders, named after its title: Videos, Pictures, Music or Documents, in Morphy\<automation name>. The automation's ⋯ menu changes that: Choose another folder…, Back to the usual folder, or Don't copy them.
Your plan, your computer
- Plan limits: when Claude's or Codex's plan passes 85% of its 5-hour or weekly limit, a run's new chats wait ("Paused · Claude at 91%") and start again when it resets. Chats already working carry on. Run anyway starts them now.
- Long jobs: a chat that leaves work running in the background, like a video render, keeps its row working until it reports back. If you quit Morphy mid-run, those rows carry on by themselves the next time you open it. Stop on such a row ends that work; when the row's output is already there (a page with a dev server still running), it's kept. Work that is still running after 3 hours is ended the same way.
- Disk space: when a run ends, Morphy deletes what its chats made along the way and keeps only the outputs, and any folder an output points into (a page's
assets/). ⋯ → Storage and clean up… says how much an automation takes and cleans up older runs. - Notifications: when outputs wait for you and Morphy isn't in front, your computer says so. It also says when a row has gone quiet for 10 minutes.
Change it
Tell the conductor. It edits the automation's flow file, automation.json, and checks it. A run keeps the steps it started with, and the next run uses the new ones. You can also open the file yourself from the ⋯ menu (Open the flow file). If the conductor's card is gone, ⋯ → Bring back the conductor brings it back.
The conductor only changes its own automation. If you ask for something automations can't do yet, it says so and offers the closest thing they can do.
Templates
| Template | What it makes |
|---|---|
| Cinematic reels | Short vertical videos in a retro low-poly 3D style: two characters, one short conversation, voices and subtitles. It writes five scripts a run, and each one you keep becomes a video. Its conductor asks what the reels are about, who the two characters are, the look and the length. The first video builds the scene, and every later one reuses it so the series stays consistent. It needs Claude, an ElevenLabs key for the voices, ffmpeg and Chrome or Edge. |
Where things are
Everything an automation uses lives in your data folder:
automations\<id>\: its flow file, the conductor's folder and instructions, and every run.automation-runs\<id>\: what each chat made, one folder per row.
⋯ → Delete automation… deletes its flow, every run and everything its chats made, and closes its conductor. Outputs you approved stay where they were copied.