VP, Product and Engineering
Changelog
This week in Adapt: browser actions, personal knowledge, and branded workspaces

We're back for week two of our weekly Adapt roundup. Last week I introduced Adapt and shared why we're writing these by hand. The short version is simple: we're building the integrated coworker that connects to your company's systems, data, and processes, and we want to build it in the open.
This week is a good example of what that means in practice. Adapt can now move from understanding the page in front of you to taking action in it. It can remember the way you work without teaching those preferences to everyone else. And your workspace can look a little more like your company, not ours.
Let's get to it. What shipped this week?
TL;DR
For the skimmers among us:
- Browser actions from the Chrome extension. Adapt can now reread a page, inspect specific parts of it, fill fields, and click controls in the tab where you started the request. You still approve clicks by default.
- Personal knowledge. Adapt now separates your personal preferences, writing style, templates, and working conventions from shared company knowledge automatically.
- Organization logos. Admins can upload a company logo that appears in the workspace menu, making Adapt feel more like part of your company.
- Quality of life improvements. We cleaned up navigation, improved settings on mobile, preserved voice dictation in chat drafts, and made long tables easier to read on smaller screens.
Now to go deeper.
Take action in your browser
The original version of our Chrome extension gave Adapt context from the page you were viewing. That alone is useful. You could summarize a long document, research a prospect while looking at their website, or draft a response using the page alongside data from your connected tools.
But understanding a page is only half the job and useful but not amazing. We're not content with not amazing!
With Chrome Extension v0.2.0, Adapt can now reread the current page as it changes, inspect a specific form or section, fill supported fields, and click controls in the same tab. You can ask it to do things like:
- "Fill this form with the company details from HubSpot."
- "Search for this customer and open their billing page."
- "Open the first matching result and summarize what changed."
The part I care about here is the handoff between context and action. Adapt already knows how to pull live information from your integrations and company knowledge. The extension gives it the page you are working on. Browser actions connect the two, so the result does not have to end as a paragraph telling you what to do next.
You stay in control. Reading the page and filling fields can happen automatically, but clicks pause for approval by default. You can deny the action, allow it once, or always allow future clicks on that site in your current browser profile.
Browser actions stay attached to the tab where the request started, so Adapt does not wander into a different tab while you work. Keep that tab and the side panel open while it finishes.
Get the Adapt Chrome extension →
Knowledge that is actually personal
Adapt's knowledge base is how it carries useful context from one conversation into the next. It remembers durable company facts, definitions, procedures, and preferences so you do not have to start every chat by explaining how your business works.
The problem is that not every useful piece of context belongs to the whole company.
My e-mail style is a good example. I want Adapt to remember that I write warmly, keep things direct, and sign off with "Dustin." I do not want everyone at Adapt suddenly signing their e-mails with my name (although frankly it would be funny once).
Adapt now gives every knowledge entry one of two scopes:
- Organization for shared facts, procedures, definitions, vendors, and team conventions.
- Personal for one person's preferences, writing style, templates, and working conventions.
Adapt chooses the scope automatically as it learns. Say, "I like reports as short bullet points with the answer first," and that preference follows you without cluttering the organization's shared knowledge base. Ask Adapt to learn your writing voice from your sent e-mails or public posts, and the resulting style guide stays yours.
Personal knowledge is visible and retrieved only for that person. Organization knowledge remains available to the team. Both are searched together when you ask a question, so there is no extra mode to remember or label to add.
You can browse, edit, or remove either scope from Settings → Knowledge.
Read more about personal knowledge →
Make the workspace yours
This one is smaller, but it matters. Adapt is joining your company as an integrated coworker. It should feel like it belongs to your company too.
Organization admins can now upload a company logo from Settings → Organization → General. Once it is saved, the logo appears in the workspace menu in place of the default Adapt mark. Wide wordmarks, square marks, and transparent images are all resized to fit without flattening them onto an awkward white box.
It is a simple bit of customization. It also makes a shared Adapt workspace much easier to recognize when you belong to more than one organization.
Quality of life improvements
A few smaller improvements shipped alongside the larger releases:
- Cleaner navigation. We split organization and personal settings into clearer menus, moved Skills and Support out of the main sidebar, and polished menu transitions.
- Better mobile settings. Settings pages now behave more predictably at smaller breakpoints.
- Voice drafts persist. Text added through voice dictation now survives in the same chat draft as typed text.
- Readable tables on mobile. Wide Markdown tables scroll horizontally instead of breaking the page.
- More resilient active chats. The composer stays locked and the working state stays visible across a refresh or deploy, so an in-progress response does not look like it stopped.
That's it for week two. Thanks for following along. You can watch the changelog for updates as they ship, and I'll be back next Friday with the roundup.
You might also like
Agents need a shared filesystem
Why persistent, organization-wide storage is a foundational primitive for AI agents that need to build on one another's work.
How to build a natural-language agent
The trick is not writing more code. It is teaching your agent the way you would teach a new teammate: show it the work with full context, write down how you do it, then let it do real work.
The Agentic Loop: Building a Software Factory Worth Its Tokens
An agentic loop is an LLM that works in a loop until the job is done. How those loops stack into a software factory, and how to keep one worth its tokens.

Make your company
instantly AI native.
Bring the integrated coworker to your whole team. Get started free with $100 in credits when you add Adapt to Slack.