---
id: LIB-3
title: The streamline editor
group: The Library
summary: One editor for every streamline, laid out from the published format — sections, briefs and instructions as a page, problems where they are
keywords: [editor, streamline editor, suggestions, suggestion chips, rules, shows when, let the model suggest, filter, chips, sections, find a section, schema, fields, more, instructions, markdown, problems, check, save, stage, what people see, what the assistant reads]
---

Open a streamline in the **Library** to edit it. The editor is laid out from Streamliner's published format, so every setting it knows has a place, with a line under each saying what it is for.

## Sections

- **Streamline**: **Name**, **Description**, **Icon**, and the **Theme**. The **Icon** row shows what the streamline wears (its logo, or its symbol in its color) and offers: **Choose image…** (Files, or drop an image on it on the Mac), **Photos…** (iPhone and iPad), **Use <site>'s logo** (the service's own logo, fetched once from the site of the streamline's first server), **Symbol and color…** and, with a logo set, **Remove logo**; a quiet line then names the fallback symbol. Images are cropped square and kept small. **Symbol and color…** opens the picker: the **Color** comes first (one for light, one for dark; it draws the icon, your messages and the accent when no theme is set), then every icon in groups, with a search that understands plain words (*chat*, *money*, *delivery*, *customer*), best matches first; any SF Symbol name typed in full works too.
- **Save** (and **Revert**, once something changed) are in the header, in the same round buttons as the header's others. ⌘S saves on a keyboard; ⌘[ goes back.
- **Instructions**: the **Briefs** it includes (LIB-2), then its own instructions, readable as a page (**View**) or editable with a formatting bar (**Edit**).
- **Model**, **MCP servers** (each server's address, sign-in and tool rules; see FR-3 and FR-4), **Welcome and quick phrases**, **Suggestions**, **Cards**.
- **Suggestions**: the chips above the composer after the first message. **Rules** lists them: each has a **Label**, one thing a tap does (sends words, puts words in the composer, opens the scanner, or runs a tool your server lets a button run) and a **When** that says when it shows, read from the chat's data (*the last step looked up the account*, *a balance is due*). The first three that hold show, in this order, and they change as the data does. **Shows when…** above the list draws the row you would see, **With sample data** or with **Nothing known yet**, and marks each rule: shows, doesn't hold, or left out (and why: a fourth one, a label too long or reading a value that isn't there). The values it assumed are listed under it. On a streamline of your own, **Let the model suggest** lets the assistant offer up to three short replies of its own when no rule holds: words only, never a tool, a few tokens a turn on your key.
- **More**: every other setting the format has. A setting a newer version of the format adds appears here without an app update.

Settings that are optional say so. Lists have **Add**, and their items can be moved or removed.

## Finding a section

A row of chips sits above the form: **All**, then one for each section (**Streamline**, **Instructions**, **Model**, … **More**). Tap a chip to show only that section; tap more chips to show several together, and tap one again to hide it. **All** shows everything again. A chip with a red count has problems in its section, so a problem is never out of sight. The editor remembers your choice while the app is open, separately for a streamline of your own, **What people see** and **What the assistant reads**. On an iPhone the row scrolls sideways.

## Problems

Problems show under the setting they concern: a color that isn't one, a required field left empty, a brief that doesn't exist, a tool rule naming a tool twice. The editor uses the same check as the Concierge and the command-line tools, so what it accepts will work.

## An organization's streamline

Two tabs: **What people see** (name, look, who may open it, welcome and composer, suggestions, cards) and **What the assistant reads** (briefs and instructions, model, MCP servers and tools, limits and network). **Stage…** in the header shows the changes (*Stage these changes?*) and **Stage** keeps them on this device until they are published (LIB-1, AD-5). The second tab is filled in once an owner has approved this device for publishing; before that it says the private half isn't on this device.

An owner whose device is approved also sees **Move to …** (the organization's name) under a streamline of their own on this device: it stages a copy as the organization's streamline, using the organization's model key instead of yours, and the copy here stays.

## Saving

On this device, **Save** keeps your changes, and **Revert** goes back. A change that would let a tool run without your tap (a suggestion that runs a tool counts), and a new streamline, shows its differences first and waits for **Save** on that card. Under the form, on this device: the streamline's **Own model key (optional)**, **Share as a file (never keys or sign-ins)** and **Delete this streamline**.
