The ops lead who's back to hands-on development
Simon works for a digital agency, heads up operations, and — after a stretch doing other things — is going back into development, "purely doing everything with AI," mostly Claude. He'd already tried the obvious tools before landing on xTiles: Notion felt too complicated, and Markdown in Obsidian was fine at first but "gets too messy" over time.
His plan wasn't narrow. He wanted xTiles for agency and client work, for personal projects like fitness tracking, and for side projects — and he was already thinking about embedding Jira tickets directly into it, on the theory that it should be possible with a simple embed.
But the one thing he came into the call specifically wanting to understand was narrower: how to wire Claude into xTiles through MCP, and use it as his daily planner.
Wiring Claude into xTiles
The setup itself is a short, one-time process:
Open Integrations in your xTiles profile
This is where the connection to Claude (or ChatGPT) lives.
Create an API token
One token, generated once, is what authorizes the connection. Access is private by default — it isn't automatically shared across a whole team, though an admin can grant teammates access separately.
Paste it into Claude's developer settings
Add the token where Claude expects an MCP server config. The exact screen depends on your OS, but the content you're pasting is minimal — essentially just your account and the token, nothing more to configure.
That's the whole thing. As Simon put it once he'd seen it: "super easy to set up."
Not just an output destination
Most people set up an integration like this expecting one direction: ask Claude to do something, and it writes the result into xTiles. That's real and useful — but it's only half of what Simon was after.
The part he specifically wanted was the other direction: xTiles as context Claude reads from, not just a destination it writes to. A project he'd built — in his case, one tracking a Jira integration — could sit there as background Claude already knows about, instead of having to be re-explained at the start of every conversation.
Right now, MCP write access is focused on tasks and planner content — Claude can create and update those directly. Full in-place editing of arbitrary project content is on the roadmap, which is worth knowing going in: the context-reading side already works well; the write side is still growing.
What a connected daily planner looks like
The planner is the part Simon was pointed toward first, and for good reason: every task from every project shows up there automatically, dated and ready — including tasks a teammate assigned to you, which surface the same way as ones you created yourself. Google Calendar events sit alongside them, and everything is draggable across days without touching the source project.
On top of that structure, the daily page can be populated automatically by Claude on a recurring prompt: a pass through the inbox, a check of xTiles tasks, a look at Slack for threads worth attention, and a summary of the inbox from roughly the last 24 hours — with the original emails archived once they're summarized. Run daily, weekly, or monthly, it turns "open five apps to find out what's going on" into one page that's already built by the time you sit down.
For agency and client work specifically
As an ops lead, the collaboration layer mattered as much as the AI layer. Tasks live in a shared collection across a team's Spaces — reassign a task to a teammate and it moves into their planner the same way it appeared in yours. Opening a task gives room to embed Google Docs, Drive files, or comments directly, which covers a good chunk of what a small agency needs without leaving the page.
Simon's Jira-embed idea fits naturally here: for a team already tracking engineering tickets in Jira, embedding those tickets into the same project that holds client context and planning is a reasonable next step once the basics are in place — client work, side projects, and personal tracking, all inside the same structure rather than three different tools.
A few ways AI speeds up the setup itself
A handful of smaller features came up that are worth knowing about on day one, since they cut down the blank-page problem that made Notion feel like too much work in the first place:
- AI-drafted templates. If nothing in the template gallery fits, ask the built-in AI assistant for a starting layout in plain language — a quick plan, a tracker, whatever the shape is — and regenerate it if the first pass misses.
- Create a project from a link. Paste a URL and xTiles builds a condensed, one-screen summary project from it — useful for turning a long article into something scannable instead of scrolling through the original.
- Web Clipper for anything you're browsing. The extension captures a page as you go — save it to a project, turn it into a task, or generate a quick summary — without breaking from what you're reading.
- Library for everything you've saved. Every link captured across your planner lands in one dedicated view, separate from the task-based planner view, so saved articles and references don't have to be hunted down project by project.