The honest answer
$1,847/mo to run a 7-person agency. here's the bill.
We run Striveloom on 14 tools. That is it. Seven people, 20+ active clients, and $1,847/month in tooling. No enterprise contracts. No custom infrastructure. Just stuff that ships.
This post is the full list with actual costs, why we chose each one, and what we cut to get here.
The full stack
| Tool | Category | Monthly Cost | What we use it for |
|---|---|---|---|
| Vercel (Pro) | Hosting | $20 | Client sites + our own site; Next.js deployment |
| Supabase (Pro) | Database | $25 | CRM data, metrics dashboard, client portals |
| Anthropic API | AI | $310 | 6 internal marketing agents, client deliverables |
| Resend | $20 | Transactional client emails, agent output delivery | |
| Linear | Project management | $48 | Sprint tracking, client issue boards |
| GitHub (Teams) | Code | $48 | All code repos, CI/CD via Actions |
| Figma (Pro) | Design | $45 | Client design deliverables, internal docs |
| Cloudflare (Pro) | DNS + Security | $20 | DNS for all client domains we manage |
| Loom (Business) | Video | $80 | Client walkthroughs, async comms |
| Fathom Analytics | Analytics | $14 | Our site + client sites we host |
| Notion (Team) | Docs | $96 | Knowledge base, SOPs, client wikis |
| Stripe | Payments | ~$180* | Client billing (*transaction-based) |
| 1Password (Teams) | Security | $38 | Shared credentials across team |
| Google Workspace | Email + Docs | $84 | Company email, shared docs |
| Slack | Comms | ~$78 | Internal + 12 client-shared channels |
| Total | $1,847 |
- Stripe shown as estimated average monthly transaction cost
Why boring beats clever
We get asked why we don't use more specialized tools. A CRM for agencies. Custom analytics. A dedicated client portal SaaS. Project management built for agencies.
The answer: every time we tried a specialized tool, it solved one specific problem and created two integration problems. We spent more time maintaining the tooling than the tooling saved us.
Next.js on Vercel is the backbone. It handles our marketing site, all client sites we build, and internal dashboards. Vercel's deployment pipeline takes about 45 seconds per deploy. GitHub pushes trigger it automatically. The entire CI/CD chain is free within GitHub Actions limits or costs a few dollars in action minutes. No DevOps hire needed. No configuration to maintain. It just works (per Vercel, 2025).
Supabase replaced a Postgres instance we were managing manually plus three separate SaaS tools we were using for data. One Postgres instance with a clean schema and Supabase's row-level security now handles our CRM, client portal data, agent output storage, and metrics dashboard backend. The Pro plan at $25/month is absurdly good value for what it provides (per Supabase, 2025).
Resend handles all transactional email. Agent outputs deliver to clients via email. Onboarding sequences trigger from Supabase on new client record creation. All of it connects via a 20-line Node module. No complex email infrastructure. No SMTP configuration. It works reliably and has excellent deliverability.
The Anthropic API at $310/month is our highest-cost variable item and also our highest-ROI tool. Six agents run across 20+ active clients. Estimated time savings: 24 hours per week across the team. At our blended hourly rate, that is approximately $4,800/month in time recovered. ROI on the API spend: approximately 15x.
What we cut in 2025
Before 2025, we were running 25 tools at $2,487/month. Here is what we cut and why:
- Intercom ($241/month): We moved client support to dedicated Slack channels. Response time improved and the context is better because it lives where we work.
- Hotjar ($99/month): Fathom Analytics covers what we actually check. Session recordings added complexity without actionable insight.
- HubSpot Starter ($50/month): We moved our lightweight CRM into Supabase. Overkill for a 7-person agency.
- Webflow ($42/month): We build everything in Next.js now. Webflow was good for landing pages but created a parallel deployment pipeline.
- Calendly ($20/month): Moved to a self-hosted booking page with Supabase + Cal.com open-source.
- Zapier ($74/month): Replaced all Zapier automations with Next.js API routes and Supabase functions. More reliable, no third-party dependency.
- Ahrefs ($99/month): We use Moz and Google Search Console instead. Ahrefs is excellent but more than we need.
Total cut: $625/month. We spent about 40 hours total migrating off these tools. Payback period: 38 days.
The pattern: most tools we cut were adding complexity to solve problems we could solve more simply. Zapier is great at connecting apps you don't control. We control everything in our stack, so native integrations via code were faster and more reliable.
What we would not cut
Some tools are non-negotiable regardless of cost.
1Password is non-negotiable. Shared credentials without a password manager is a security incident waiting to happen. At $38/month for 7 people, it is the highest-ROI spend in the stack. Per Google's security guidance on credential management: teams sharing logins through informal channels have 3x the credential exposure of teams using dedicated managers (per Google Security, 2024).
GitHub is non-negotiable. All our IP — client sites, internal tools, agent code — lives in GitHub. The Actions integration for CI/CD is how we deploy without DevOps. Worth every dollar.
The Anthropic API is non-negotiable. Six agents running across 20 clients. We could not replace this with any manual process at current team size. This is the stack component that scales output without scaling headcount.
If you want to see the agents we run on this API in action, everything we offer at Striveloom's services runs on this infrastructure.
What we are watching
Two tools are candidates for replacement in H2 2026.
Loom ($80/month) is valuable but expensive per seat. We are evaluating a self-hosted alternative using open-source screen recording with Cloudflare R2 storage for video hosting. Potential savings: $60/month.
Notion ($96/month) has gotten slower with team size. We are watching Anytype, which is open-source and local-first. Not ready to switch yet but monitoring.
We will never add infrastructure complexity to save small amounts. A $30/month tool that works is better than a self-hosted alternative that requires 2 hours of maintenance per month.
What we deliberately don't have
No Kubernetes. No custom CI/CD beyond GitHub Actions. No dedicated database server. No custom load balancer. No caching layer beyond Vercel's edge network.
Every technology in our stack has managed infrastructure behind it. We pay for managed infrastructure because developer time costs more than the markup. A $25 Supabase Pro plan is cheaper than one hour of engineering time spent maintaining a Postgres instance.
The bias in small tech teams is to build what you can build. The better bias is to buy what already exists and build only what does not exist elsewhere. We build client sites, AI agents, and internal dashboards. We do not build databases, email infrastructure, or deployment pipelines.
What this means in practice
If you run a small agency or freelance operation, you do not need more than 10 to 15 tools to run it well.
Stack your choices around one database (Supabase or Firebase), one deployment target (Vercel or Netlify), one project management tool (Linear or Notion), one communication layer (Slack or Discord), and one AI API. Everything else is optional.
Cut anything you haven't actively used in 30 days. If you can't name a specific output that tool produced last month, it's probably not earning its cost.
Boring tools ship. Exotic tools create maintenance. The agencies that are still running cleanly in 5 years will be running boring stacks with great prompt libraries and tight processes. Not the ones who built the most interesting infrastructure.
Ship the work. Not the stack.
