Platforms & Tools
Platforms and Tools
A curated list of best-of-breed platforms and tools to consider for implementation.
We utilize a variety of technologies selecting the optimal stack for each project.
Open Source Radar
How we decide what is worth adopting
We track open source projects by how fast they are actually growing, measured from daily GitHub event data over real calendar windows — not total stars, which mostly measures how long a project has existed.
AI and agentic engineering
How we build now. These are production tools in our own delivery, not a lab experiment running beside it.
Claude Code
Anthropic's coding agent, running in the terminal, the editor, and unattended in automation. It reads and changes a codebase directly instead of suggesting snippets to paste, which is what makes it useful on work that spans many files.
Cursor
An AI-native editor with background agents that take a scoped task and open a pull request. We use it for parallel work: several agents on independent slices while a human reviews the diffs.
OpenAI Codex
OpenAI's software engineering agent, available as a CLI and a cloud sandbox. A second independent model earns its place on review and audit work, where agreement between two of them is a stronger signal than either alone.
Ollama
Runs open-weight models on your own hardware. This is the answer when the data cannot leave the building: the prompt, the source, and the output never reach a third-party API.
Model Context Protocol
An open standard for connecting AI assistants to tools and data. One integration works with every client that speaks it, so the connector you build is not locked to the assistant you built it for.
Agno
A Python framework for multi-agent systems, with typed tools, memory, and workflow control. Suited to agents that have to run as a deployable service rather than as a chat session.
Cloud and infrastructure
Where it runs, and how the running of it is kept reproducible.
Cloudflare
Workers, Pages, object storage, and a global network in one platform. Code runs close to users with no regions to choose and no servers to patch. This site is built and served on it.
Convex
A reactive backend in TypeScript: database, server functions, and live subscriptions with no separate API layer to keep in sync. Its per-branch preview deployments let a database change ship behind the same review as the code.
OpenTofu
Infrastructure as code, and an open source fork of Terraform. Native state encryption is the reason we prefer it: deployment credentials end up in state, and Terraform writes those in plaintext.
Docker
Containers, so the thing that passed on a laptop is the thing that runs in production, and a new contributor gets a working environment from one command instead of a setup document.
GitHub
Source, review history, and issues. Pull requests keep the reasoning for a change next to the change itself, which is most of what makes a codebase legible to whoever inherits it.
Frameworks and languages
The stack we reach for first, and the ones we support where a client already stands.
TypeScript
A typed superset of JavaScript, and the default for everything we write on the web. Types are worth more now than they were two years ago: they are the contract an AI agent has to work against.
Next.js
A React framework covering static export, server rendering, and routing in one place. Static output is what lets a marketing site be fast and cheap without giving up a real component model.
React
The component model underneath most of what we build for the web, and the one with the deepest hiring pool — which matters when a client has to maintain the result.
Tailwind CSS
Utility-first styling with design tokens as the single source of truth. A brand change lands in one file instead of being chased through a stylesheet.
Expo
React Native with the build, signing, and over-the-air update pipeline already solved. One codebase reaches iOS and Android, and a fix ships without waiting on app review.
Flutter
Natively compiled apps for mobile, web, and desktop from a single codebase. The right choice over React Native when the interface is heavily custom-drawn rather than assembled from platform controls.
Angular
A batteries-included framework from Google, with structure imposed rather than assembled. Common in enterprise estates, and where we meet a lot of existing code.
Python
The language of the data and AI layer. Every serious model, agent, and pipeline library lands here first, which is why our backend work for AI systems starts in Python rather than ending up there.
Drizzle ORM
A TypeScript ORM that stays close to SQL instead of hiding it. Queries are typed end to end, and the generated SQL is the SQL you would have written.
Ethers
A library for interacting with Ethereum: contracts, wallets, and transactions. We use it where a client has a genuine settlement or provenance requirement, not as a default.
Delivery and design tooling
What proves the work is right before anyone else sees it.
Playwright
Browser tests that drive a real browser. This is the check that stops an agent-written change from shipping on the strength of it looking plausible.
Visual Studio Code
The editor most of this happens in, and the extension surface that the current generation of AI coding tools is built on top of.
Zeplin
A handoff point between design and engineering, carrying specs, assets, and component states so that spacing and type are read rather than guessed at.
Salesforce Platform
CRM plus the platform layer around it: data model, automation, and integration. Where a client already runs their business on Salesforce, we build alongside it instead of next to it.
Android
Google's mobile operating system, and the wider device reach of the two platforms. Its fragmentation is the reason real-device testing stays in scope rather than being cut.