Files
Files is your organization’s shared workspace — the documents Adlyse works from, as opposed to the numbers it reports on. Briefs, playbooks, house rules, saved research, anything you’d otherwise have to re-explain in every conversation. Agents with folder access read and write here too, so it’s the common ground between you and them.
In one line
Anything you find yourself pasting into a chat more than twice belongs in Files.
Getting things in#
Open Files from the sidebar. You can create folders, create markdown files directly, upload from your machine (drag a folder in and its structure comes with it), and move things around by dragging. Files are versioned, so an edit is never destructive — you can look at earlier versions and promote one back.
Brand assets and logos#
Where the Creative Studio is on, brand assets live there rather than in Files: upload logos, product images and brand elements in the Creative Studio’s Brand tab, or use Move to Creative Studio on an image in Files. Agents still see them, read-only, under /org/brand/logos/, /org/brand/products/ and /org/brand/elements/, and the old path of a moved file keeps working, so nothing that pointed at it breaks.
Without the Studio, the brand/ folder plays that part: Adlyse automatically uses images there to ground on-brand creative generation. Put logos in brand/logos/. In an agent’s workspace the same folder appears as /org/brand/, so agents and the creative service discover it without you attaching the logo in every chat. Existing logo/ and logos/ folders also work: Adlyse checks them when brand/ has no usable logo.
If you name a specific logo or attach one for a creative, that choice takes precedence over automatic selection and keeps the exact variant you chose.
Use transparent PNG or WebP files and name each lockup for what it is and where it belongs. The name is how Adlyse distinguishes formats and chooses a visible logo for the generated background:
| Example | Meaning |
|---|---|
brand/logos/primary-color.png | Default full-color logo |
brand/logos/wordmark-white.png | Wordmark for dark backgrounds |
brand/logos/wordmark-black.png | Wordmark for light backgrounds |
brand/logos/icon-color.png | Compact icon or brand mark |
Words such as primary, wordmark, and icon identify the lockup. white or reverse means the file is intended for a dark background; black or dark means it is intended for a light background. Transparent files without a background-specific name are treated as suitable for either. Keep equivalent light and dark versions in the same format when you have them—for example, two wordmarks—so a background change does not unexpectedly swap a wordmark for an icon.
For a title card, report header, or chat avatar, specify where the official logo belongs. Adlyse can reserve those positions and place the original logo after generating the scene, preserving its proportions and transparency. Review the result for placement and unintended extra marks before using it.
Use raster logo files
SVG uploads are rejected for security. Export the original logo as a high-resolution transparent PNG or WebP; Adlyse preserves those pixels when it places the logo rather than asking the image model to redraw it.
Guidelines: standing instructions#
The guidelines/ folder is special. A markdown file in there becomes a standing instruction that Adlyse follows on every message — your do’s and don’ts, escalation rules, and platform playbooks. There’s no separate step to switch one on; create the file and it’s picked up on the next message.
Every organization starts with one: guidelines/adlyse.md, covering how much the assistant may do on its own, what it must confirm first, and what it must never touch. It’s a starting point, not a fixed policy — edit it to match how your team actually operates.
The filename decides when a guideline applies. Name one after an ad platform and it only applies when that platform is connected:
| File | Applies |
|---|---|
guidelines/adlyse.md | Always |
guidelines/google.md | Only when a Google Ads account is connected |
guidelines/meta.md | Only when a Meta account is connected |
guidelines/microsoft.md | Only when a Microsoft Ads account is connected |
guidelines/chatgpt.md | Only when a ChatGPT Ads account is connected |
guidelines/anything-else.md | Always |
So a Google rule never bleeds into a Meta conversation, and a playbook for a platform you haven’t connected yet costs nothing to write early. Any other name — creative-review.md, client-reporting.md — is a house rule that always applies.
Platform playbooks aren't pre-written
Only adlyse.md ships with starter content. Bid strategies, naming conventions, and creative standards differ far too much between accounts for a default to be right, and a rule nobody read is worse than no rule. Write the ones you want.
There is no brand guideline file
Brand context used to live in a guidelines/brand.md file. It doesn’t any more — it comes from your Brand profile, which is rendered live into every conversation. Maintain your positioning there, in one place, rather than in a document that can drift out of date. Any leftover brand.md from before the change is archived and ignored.
Writing a good guideline#
Write them the way you’d brief a new teammate — short, direct, specific. The ones that earn their keep tend to be the ones that encode a decision you’d otherwise have to make again:
Never propose changes to brand-term campaigns without flagging it explicitly.
Report in plain language — assume the reader is a founder, not a media buyer.
Escalate anything over $2,000/mo of impact to a human before proposing it.Changes take effect on the next message.
Files and agents#
An agent with the Folder access capability can browse and edit this workspace — read your briefs before writing a report, and file its own output where your team will find it. It’s one of the more useful capabilities to turn on, and one of the few where “can write” is genuinely what you want.
See tool groups for the details.
Common flows#
- Stop repeating yourself. Write the context you keep pasting into chats as a house-rule guideline once.
- Codify a platform playbook. Put your Meta rules in
guidelines/meta.mdand they’ll stay out of Google conversations. - Give agents somewhere to file things. Point a scheduled reporting agent at a folder and let it deposit each run’s output.
Positioning, audience, and objectives — the context, not the rules.
TablesStructured rows, for when a document isn’t the right shape.
SkillsA guideline applies always; a skill applies to one kind of task.
Tool groupsGive an agent folder access — and understand what that grants.
Read and write the same workspace with list_files, read_file, write_file, and grep_files; guidelines are also readable via list_guidelines and get_guideline.