Skip to content

AI consulting, production builds, real systems.

Inspired By Frustration is a one-operator studio shipping production backends, typed APIs, MCP servers, AI agents and full-stack web apps. Twelve years of engineering taste, AI-accelerated delivery, and a refusal to ship the kind of software that made you hate software in the first place.

Inspired By Frustration is an AI consulting and AI-accelerated development shop. We ship production AI agents, MCP servers, typed APIs, and full-stack web apps — with fractional CTO advisory included. One operator orchestrating a swarm of coding agents, twelve years of engineering taste, zero handoffs between schema, server, and client.

now buildingMCP server · billing agent
stackPostgres · TS · edge
reply within< 24h · operator-direct
last ship11h 00m ago
local · sthlm--:--:-- UTC

// stack · what I actually reach for

One brain, top to bottom.

Six layers, no handoffs. Swap any item for the one your team already runs — the shape of the work doesn't change.

missing something? ask →
  1. /01

    Language & runtime

    typed, boring, fast

    • TypeScript
    • Node 22
    • Bun
    • Python 3.12
    • Go (when it earns it)
  2. /02

    Data & persistence

    Postgres or it didn't happen

    • Postgres
    • Drizzle
    • Supabase
    • pgvector
    • Redis
    • S3 / R2
  3. /03

    Backend & APIs

    edge-first, queue-aware

    • Hono
    • tRPC
    • REST
    • Cloudflare Workers
    • Inngest
    • Stripe
  4. /04

    Agents & MCP

    the bit nobody wants to own

    • MCP servers
    • Claude
    • GPT-5
    • LangGraph
    • OAuth 2.1
    • OpenTelemetry
  5. /05

    Frontend

    only when the brief forces it

    • React 19
    • Next 16
    • Vite
    • Tailwind
    • shadcn/ui
    • Framer Motion
  6. /06

    Build & ship

    the swarm + the senior

    • Lovable
    • Cursor
    • Claude Code
    • Vercel
    • Cloudflare
    • GitHub Actions
postgres·
typescript·
react 19·
vite·
supabase·
fly.io·
cloudflare workers·
hono·
drizzle·
openapi·
zod·
mcp·
openai·
anthropic·
langgraph·
pgvector·
inngest·
stripe·
playwright·
vitest·
json-ld / seo·
tailwind v4·
postgres·
typescript·
react 19·
vite·
supabase·
fly.io·
cloudflare workers·
hono·
drizzle·
openapi·
zod·
mcp·
openai·
anthropic·
langgraph·
pgvector·
inngest·
stripe·
playwright·
vitest·
json-ld / seo·
tailwind v4·
postgres·
typescript·
react 19·
vite·
supabase·
fly.io·
cloudflare workers·
hono·
drizzle·
openapi·
zod·
mcp·
openai·
anthropic·
langgraph·
pgvector·
inngest·
stripe·
playwright·
vitest·
json-ld / seo·
tailwind v4·

/ first reply

< 4h

if you need help now

/ working prototype

72h

end-to-end, not a mockup

/ ship window

2–6 wk

to production, not staging

/ team size

small

senior core + bench on demand

[ 01 / services ]

Hire the operator,
not the agency.

// no PMs. no decks.
direct line to the person writing the code.

/ 01

Backends & APIs

Typed end-to-end. Postgres, Hono/tRPC, queues, billing, auth. Built so the next engineer thanks you, not curses you.

PostgresTypeScriptHonoDrizzleCloudflareStripe
/ 02

MCP Servers

Model Context Protocol servers your agents can actually trust. Tools, resources, OAuth, rate limits, OpenTelemetry traces.

MCPNodePythonOAuth 2.1OTelJSON-RPC
/ 03

AI Agents

Production agents that don't hallucinate themselves into a corner. Evals, guardrails, fallbacks, cost ceilings, replay.

ClaudeGPTLangGraphInngestpgvectorBraintrust
/ 04

AI-Accelerated Builds

Full products in 2–6 weeks. Lovable + Cursor + 12 years of taste. Frontend only when the brief forces it.

Next 16RSCEdgeSupabaseVercelResend

[ 02 / how it ships ]

99 PRs. 20 hours. Zero drama.

A swarm of 6–10 agents, one senior operator. Every PR is typed, scoped, evaluated and runbook-backed before it touches prod. You watch the feed. You ship by Friday.

118
PRs / day
38ms
p95 latency
0
escapes 90d
live · agent-feed.logstreaming
14:02:11agent-03 · branch feat/pricing-quote opened
14:02:14PR #412 · typed tool scaffolded · scope=pricing:read
14:02:19evals · 142/142 ✓ · p95 38ms · 0 escapes (90d)
14:02:22merged · runbook attached · deploy queued
14:02:27agent-06 · branch fix/rate-limit opened
14:02:31PR #413 · ctx.rateLimit("quote", 60) · diff +4 -1
14:02:34merged · prod green · users unaffected
14:02:38queue · 7 PRs in flight · 6 agents active
14:02:41
main · clean6 agents · 7 in flight✓ green

[ 04 / how it goes ]

The dance,
described honestly.

// four moves. no choreographer. usually fun.

Every good piece of software I've ever shipped started as someone being annoyed enough to fix it properly.
— that's the brief. that's the whole brief.
  1. 1You email me. I reply. We're on a call within four hours if you're on fire — otherwise inside the day. No discovery deck, no junior account manager, no procurement ballet shaped like a Gantt chart.
  2. 2You'll say you need it yesterday. I'll say that's physically impossible. Then I'll quietly over-deliver on day three and pretend it was hard. You know how it goes.
  3. 3v0.1 in hours — depending on contracts and how fast your legal team drinks coffee. A serious demo / prototype in days. Then, depending on size, weeks to live. We agree on the goal, the style, a rough sketch — then the spine underneath it (API, data model, deploy path). Vibes and load-bearing parts. In that order.
  4. 4It ships. To production, not "staging we forgot about." With evals, monitoring, docs, and a handover so embarrassingly thorough your next engineer will side-eye the invoice and whisper "…did we over-pay?" (You didn't.)

[ 05 / questions ]

Common questions.

The five things buyers ask before the first call. Topic-specific deep-dives live on each service page.

  • What does Inspired By Frustration do?

    Inspired By Frustration is an AI consulting and AI-accelerated development shop run by one operator. We ship production AI agents, MCP servers, typed APIs, and full-stack web apps, with fractional CTO advisory included. Twelve years of engineering taste, a swarm of specialised coding agents under a single reviewer, and zero handoffs between schema, server, and client.

  • How does the squad-of-agents model output 15× a single developer?

    I orchestrate a swarm of specialised coding agents that cut tickets in parallel — schema, API, UI, tests, docs — while I review every diff before it lands. The output is production code, not throwaway scaffolding: typed, tested, and shipped on the same week the brief lands. Most teams see a 3–10× compression on standard work and bigger leverage on greenfield builds.

  • When should I hire a fractional CTO instead of a full-time one?

    Fractional CTO fits when you need senior technical judgment but don't yet need (or can't afford) a $250k+ executive. Typical fits: pre-Series A startups, post-PMF companies cleaning up tech debt, or AI-first teams that need someone who has shipped production AI in anger. A fractional engagement gives you architecture, hiring, and vendor calls on tap — usually 4–8 hours a week — without the equity dilution.

  • Do you only consult, or do you also build?

    Both, and they reinforce each other. The advisory work is grounded in shipping production AI agents, MCP servers, and full-stack apps every month. The build work is grounded in twelve years of engineering judgment about what to skip and what to harden. We don't do slideware-only consulting and we don't take builds we wouldn't bet our own runway on.

  • What stacks do you ship on?

    TypeScript end-to-end. Next.js (App Router) and React on the front. Node, Hono, and edge runtimes on the back. Postgres via Supabase or Drizzle. Cloudflare Workers and Fly.io for hosting. MCP servers for AI-agent integration. Stripe for billing, Playwright for E2E, Vitest for unit. Anthropic and OpenAI on the model side, with model-agnostic plumbing where it matters.

[ 06 / contact ]

Tell me what's broken.

One brief, one operator, one ship. I take 1–2 builds per month. Replies within 24h on weekdays — usually with three concrete questions and a price band.