Our Stack, Not Just Our Integrations
Reeve connects to a lot of things. Shopify, Klaviyo, Meta Ads, Google Ads — the list keeps growing. But there's a difference between "we support this" and "we bet our own business on this."
This post is about the latter. These are the tools we use every day to build and run Reeve. They're the integrations we've tested the hardest, optimized the deepest, and would recommend to anyone setting up a serious agentic workflow.
Reeve works with plenty of alternatives. But if you're starting fresh or re-evaluating your stack, this is what we'd pick — and why.
Frontend Deployment — Vercel
We ship our frontend on Vercel. It's the obvious choice for a Next.js app, but it's more than that. Preview deployments on every PR, instant rollbacks, edge functions — it removes an entire category of deployment headaches.
For agentic workflows, Vercel's speed matters. When Reeve's agents are iterating on landing pages or testing creative, sub-second deploys mean the feedback loop stays tight.
Why not alternatives? We've used Netlify and Cloudflare Pages. Both are solid. Vercel just fits our Next.js stack like a glove, and the DX is unmatched.
Backend Deployment — AWS + Railway
Our backend runs on AWS for production infrastructure and Railway for services that need to move fast.
AWS gives us the control and scale we need — EC2, Lambda, SQS, the whole playbook. Railway is where we prototype and run internal services. It's the fastest path from "I need a service" to "it's running in production," and the pricing is sane.
The split: If it needs to handle scale and we need fine-grained control, AWS. If it needs to exist by tomorrow and reliability is good enough, Railway. Both get first-class treatment in our deployment pipeline.
Database — AWS + Railway
Same philosophy extends to our data layer. We use managed databases on both AWS (RDS, DynamoDB) and Railway (Postgres).
Railway's managed Postgres is genuinely good for most workloads. When we need the durability guarantees and ecosystem of AWS, we go there. The key is that Reeve's agents can query and analyze data from either without caring about the underlying infrastructure.
Web Analytics — PostHog + GA4
We run both PostHog and Google Analytics 4. Not because we love having two analytics tools — because they serve different purposes.
PostHog is our product analytics workhorse. Session replays, feature flags, funnels, event tracking — it's where we understand how people use Reeve. It's open source, self-hostable, and the API is agent-friendly. When Reeve's analytics agents pull funnel data or identify drop-off points, PostHog is usually the source.
GA4 is the standard. Clients expect it, ad platforms integrate with it, and it handles attribution in ways PostHog doesn't try to. We use it for traffic-level analytics and as the common language when talking to clients about their numbers.
Our recommendation: Run both. PostHog for depth, GA4 for breadth. Reeve integrates with each natively.
Domain Hosting & DNS — Vercel
We keep it simple. Domains and DNS live on Vercel alongside our frontend deployment. One dashboard, automatic SSL, no context-switching between providers.
If you're already deploying on Vercel, there's no reason to manage DNS elsewhere. The integration is seamless and it eliminates an entire class of "is it a DNS issue?" debugging.
SMS — Twilio
Twilio for SMS. It's the default for a reason — reliable delivery, good international coverage, and a programmable API that agents can work with directly.
When Reeve sends order notifications, abandoned cart nudges, or support follow-ups via SMS, Twilio is the pipe. We've evaluated alternatives, and for programmatic SMS at scale, nothing else comes close on reliability.
Email — Resend
Resend is our transactional email provider. Built by developers, for developers. The API is clean, deliverability is excellent, and it plays nicely with React Email for templating.
We use Resend for system emails, notifications, and transactional flows. For marketing email (campaigns, newsletters, flows), our clients typically use Klaviyo — and Reeve integrates deeply with both. But for the emails we send as a platform, Resend is the one.
Why not SendGrid/Mailgun? We've used both. Resend is just simpler, more modern, and the deliverability has been rock solid from day one.
Workplace — Google Workspace
Google Workspace runs our company operations. Gmail, Calendar, Drive, Docs — the whole suite.
More importantly for Reeve: Google Calendar and Gmail are the connective tissue for agent scheduling and communication. When Reeve checks your calendar, drafts emails, or coordinates across team members, it's talking to Google's APIs. The integration is mature and battle-tested.
Messenger — Slack
Internal communication lives in Slack. But for us, Slack isn't just a chat app — it's an agent interface.
Reeve's agents live in Slack channels. They report, they respond, they escalate. The threading model, the API, the webhook support — Slack is built for this kind of human-agent collaboration in a way that other messengers aren't.
We also support Discord and other platforms, but Slack is where our team actually works, and it's what we recommend for any team running Reeve in a professional setting.
Project Management — Notion
Notion is our workspace for docs, specs, project tracking, and SOPs. It's flexible enough to be a wiki, a task board, and a knowledge base all at once.
For agentic workflows, Notion's API lets Reeve read and update project status, pull context from documentation, and keep track of ongoing initiatives. It's not the fastest project management tool, but its versatility makes it the right fit for a team that needs one tool to do many things.
LLM — Claude
We're all in on Claude by Anthropic. It's the brain behind Reeve's agents.
This isn't just brand loyalty — we've run extensive evaluations across models, and Claude consistently wins on the things that matter for agentic work: instruction following, nuanced reasoning, tool use, and knowing when not to do something. The extended thinking capabilities let our agents work through complex multi-step problems in ways that genuinely surprised us.
We support other models and providers. But Claude is our default, our recommendation, and the model we optimize for first.
VoIP — Quo
Quo handles our voice communications. It's a newer player in the VoIP space, but it's built with modern workflows in mind.
Clean call quality, solid API, and a team that actually ships. As Reeve's voice capabilities expand, having a VoIP provider that thinks in terms of programmable voice (not just phone lines) makes all the difference.
The Philosophy
You'll notice a pattern in our choices: developer-friendly, API-first, and opinionated about simplicity.
We don't pick tools because they have the most features. We pick them because they do their job well, expose clean APIs for agents to work with, and don't require a team of specialists to maintain.
That's the same philosophy behind Reeve itself. Your AI team should work with your existing tools, not force you to adopt new ones. But if you are choosing tools — these are the ones we'd recommend building on.
Want to see how Reeve integrates with your existing stack? Get started free or book a demo.