We like to take stock of our tools every so often. We’ve done this before for macOS productivity apps and it’s always one of our most-read posts. Partly because it’s useful to reflect on what’s actually working and partly because these posts always seem to resonate with other teams going through the same decisions.
This time around though it feels different. Our tools and ways of working have changed more in the last 18 months than they did in the five years before that. AI tooling in particular has completely reshaped how we write code. Some things we were paying for 18 months ago we’ve dropped entirely. Others we’ve been using since day one and can’t imagine replacing.
Here’s what a small UK software development team actually runs on in 2026.
Notion
Notion is our knowledge base. Every process, every playbook, every onboarding doc lives there. We’ve tried wikis, Google Docs, Confluence and various other things over the years and Notion is the one that stuck. It hits the right balance between structure and flexibility. You can spin up a quick doc without fighting templates but you can also build something more rigid when you need it.
We use it for documenting ways of working, project kickoff templates, retrospective notes and keeping track of things that don’t belong in a ticket system. It’s not perfect. Search can be slow and the mobile app is a bit flaky. But nothing else we’ve tried has come close to replacing it.
Tuple
Tuple is our pairing tool and our internal calling app. If you haven’t come across it, it’s a macOS app built specifically for pair programming. Low latency screen sharing, remote control that actually works and drawing tools for pointing at things on screen.
We moved to Tuple years ago and it’s been rock solid. When you’re pairing on a tricky React Native bug or walking someone through a Rails codebase, the quality of the connection matters. Zoom and Google Meet are fine for client calls but they’re horrible for actual development work. The lag makes everything feel sluggish and screen sharing quality drops off a cliff.
For a small team that does a lot of pairing, Tuple has been one of the best investments we’ve made.
Slack
Nothing groundbreaking here. Slack handles all our team communication and we pipe alerts into it from various services. Deploy notifications, CI failures, error tracking, uptime monitoring. The usual.
We’ve kept our channel structure pretty lean. A general channel, a channel per active project and a handful of integration channels for automated stuff. The temptation with Slack is to create a channel for everything and before you know it nobody can find anything. We try to resist that.
One thing worth mentioning is that we’ve started using Slack more for async communication than we used to. With the team spread across different schedules and some of us working with AI tools at odd hours, real-time chat isn’t always the right mode. A well-written Slack message with enough context is often better than a quick call.
GitHub
GitHub handles our repositories and CI via Actions. We’ve been on GitHub since we started the company and it’s never given us a reason to leave.
GitHub Actions has replaced everything we used to use for CI. We run tests, linting, type checking and deploys all through Actions. The workflow files live in the repo alongside the code which means CI config gets the same review process as everything else. No more logging into a separate CI dashboard to figure out why a build broke.
Linear
Linear is our project management tool. We switched from Height earlier this year and it’s been a great fit. The speed of the interface is the thing that hits you first. Everything feels instant. But what’s kept us on it is how well it maps to the way we actually work. Cycles, project grouping and the triage workflow all click without needing to fight the tool into shape.
The recent addition of MCP (Model Context Protocol) support has been genuinely useful. It means Claude Code can read and update Linear issues directly, so we can ask it to check what’s assigned to us, update ticket status or pull in context from an issue while we’re working on the code. It removes a layer of context switching that used to break flow.
We’d used Height for a long time before this and it served us well. Linear just fits better with how we think about sprints and priorities now.
Figma
Figma handles all our design work. Mockups, wireframes, prototypes and design system documentation. Our App Planning Day outputs typically include Figma files that clients can take away and iterate on.
The collaborative features are genuinely useful when you’re working with a client who wants to be involved in the design process. Being able to leave comments, suggest changes and see real-time updates removes a lot of the back-and-forth that used to happen over email with static screenshots.
We’re keeping an eye on what AI design tools are doing but nothing has replaced Figma for production design work yet. The AI features inside Figma itself are getting more useful though.
Raycast
Raycast replaced Spotlight for everyone on the team and has slowly become one of those tools you don’t think about until you use someone else’s machine and realise how much you rely on it. We wrote a deep dive on replacing Spotlight with Raycast a couple of years ago and everything in that post still holds up.
The core launcher is great but the real value is in the extensions and the AI integration. We use Raycast AI constantly. Quick code explanations, rewriting commit messages, translating error messages and all the small stuff that doesn’t warrant opening a full conversation with an AI assistant. It sits in the gap between “I need Claude to think about this properly” and “I just need a quick answer.”
We’ve also built a few custom Raycast scripts for repetitive tasks like spinning up dev environments and checking deploy statuses.
Claude Code
This is the big one and the tool that’s changed how we work more than anything else in recent memory. Claude Code is our primary coding assistant. We use it as a CLI tool alongside our editors. We’ve written about applying AI to our development workflow before but the tooling has moved on massively since then.
We went through the usual journey. Started with Copilot, tried various tools, eventually settled on Claude. The thing that made it stick wasn’t just code completion. It’s the ability to have it understand your entire codebase and make meaningful changes across multiple files. We’ve written about this before in our posts on vibe coding tools and deploying vibe coded projects.
For our day to day work on React Native apps and Rails backends, Claude Code handles everything from scaffolding new features to debugging production issues to writing tests. It’s not a replacement for knowing what you’re doing. You still need to review what it produces and make architectural decisions. But it’s genuinely made us faster and we’re shipping better code because of it.
We recently wrote about whether code smells still matter in a world where AI writes most of your code. Short answer: yes, maybe more than before.
Fastmail
Fastmail handles our email. We moved away from Google Workspace a while back and haven’t looked back. It’s fast, the spam filtering is excellent and it respects your privacy in a way that Google’s products fundamentally don’t.
The calendar works well enough, the contacts sync properly and the web interface is genuinely pleasant to use. It doesn’t have the ecosystem of integrations that Google Workspace has, which occasionally means we need a workaround for something. But for a small team that mostly just needs reliable email, it’s been brilliant.
What we’ve dropped
Worth mentioning a few things that didn’t make the cut this time:
Google Workspace was replaced by Fastmail for email. We still use Google Docs occasionally for client-facing documents where collaboration matters but it’s no longer our primary tool for anything.
Separate CI services (we used CircleCI and Bitrise at various points) were replaced entirely by GitHub Actions. One less thing to maintain and pay for.
Height was our project management tool for a long time and did the job well. We switched to Linear because it better matches how we work now, especially with the MCP integration tying it into our AI workflow.
Standalone AI chat tools have mostly been replaced by Claude Code and Raycast AI for development work. We still use Claude directly for longer research and planning tasks but the separate “AI chat window” workflow has largely disappeared from our day to day.
What we’re watching
We keep an eye on things without jumping on every new tool that launches. Right now we’re interested in:
- How AI-native development environments evolve. Tools like Cursor are interesting but we haven’t committed to switching editors yet.
- Whether Figma’s AI features get good enough to change the design workflow meaningfully.
- What happens with local AI models. Running things on-device for quick tasks without an API round-trip is appealing.
- How tech stack decisions change as AI tools get better at working with certain languages and frameworks.
We’ll probably write another one of these in a year and half the list will have changed again. That’s the reality of working in software right now. The tools are moving fast and the teams that stay curious without chasing every trend tend to do best.
If you’re building a product and want to work with a team that’s opinionated about its tools, get in touch or book a free chat.
