The tools we'd build a ai agencies & consultants on tomorrow.

How we think about the AI agency stack

Running an AI agency means doing two things well: building AI-powered systems for clients, and running your own business. Most software advice optimizes for one or the other. We do both, so we'll talk about both.

The tools in this guide are the ones we actually use. Not affiliate picks, not sponsored placements — the software on our screens every day building and delivering real AI projects.

The three layers of an AI agency stack

Layer 1: Building tools — the software you use to write code and build AI systems for clients. This is where Cursor, Claude Code, the Vercel AI SDK, and v0 live. These tools determine how fast you can build and how well the things you build work.

Layer 2: Research and thinking tools — the AI you use to think, write, research, and make decisions every day. Claude and Perplexity are in this layer. These tools multiply your individual output: better proposals, faster research, sharper thinking.

Layer 3: Operations tools — how you actually run the business. CRM, project management, invoicing, client portals. Systemly is what we built for this layer because nothing else was designed for AI agencies specifically.

What we've learned from running this stack

Don't overspend in layer 1 before you have layer 3 solved. It's tempting to buy every AI coding tool and optimize your dev environment before you have a clean way to manage clients and deliver projects. The billing mess will kill you before the coding bottleneck does.

Pick one tool per job and commit. Tool sprawl is a silent productivity killer. We use Cursor as the primary IDE, Claude as the primary chat assistant, Perplexity for research, and Systemly for operations. Fewer tools, used deeply, beats a wide collection of half-configured software.

The coding tools pay for themselves faster than you think. Cursor Pro at $20/month is a rounding error when you're billing $5,000+ per project. The question isn't whether to use AI coding tools — it's which ones, and that answer is Cursor + Claude Code.

Clients ask about your stack. When you can tell a prospective client that you build with Claude Code, deploy on Vercel, and manage everything in Systemly — that's a different conversation than "we use whatever tools." The stack is part of the pitch.

What's not in this stack

We deliberately left out a few categories:

• AI workflow automation tools (Zapier AI, Make, n8n) — useful but not core. We add these per client engagement, not as agency defaults.
• Observability and monitoring — Sentry, PostHog, Axiom. Important, but deploy-phase decisions, not stack decisions.
• Specialized vertical tools — there are good AI tools for specific industries. We cover the horizontal agency layer here; vertical tools are per-engagement decisions.

How to build this incrementally

If you're starting from scratch, add in this order:

1. Start with Claude and Perplexity (immediate daily value, no setup)
2. Add Cursor (you'll use it in the first hour)
3. Get Systemly running for client and project management (do this before you take on your second client)
4. Add Claude Code (once you're shipping code regularly)
5. Add the Vercel AI SDK when you're building production AI features
6. Add v0 when you need to prototype UIs fast