Foundation
- Use-case scoping & success metrics
- Provider selection (Claude / GPT / open-source)
- Production deployment
- Cost monitoring & rate limits
All offers
03 / Offers
AI Agent Build
An AI agent your team trusts in production. Evaluated, deployed, costs under control.
A working agent your team trusts.
Free 30-min call · No obligation
€25.000 · 4 weeks · Fixed price
A / Scope
What's included
Tools & integrations
- Custom tool definitions
- API integrations (your stack)
- Vector DB / RAG (if needed)
- Memory & context strategy
Quality & safety
- Eval suite for behavior regressions
- Prompt injection defenses
- Output validation
- Logging & traces
Aftercare
- 4 weeks bug-fix support
- Handover docs (prompts, evals, ops)
- 1 follow-up architecture call
B / Process
How it runs
Discovery first: requirements, prioritization, architecture. Then milestone-based delivery, constant communication, no surprises.
-
W1
Discovery
Use case, functional & non-functional requirements, success metrics, eval definitions. What does 'working' actually mean?
-
W2
Build I
Core agent loop, tools, integrations, first evals running.
-
W3
Build II
RAG / memory, prompt injection defenses, cost & rate limits.
-
W4
Ship
Production deploy on your accounts, full eval suite, handover docs.
C / Fit
Who this is for, and who it isn't
Who this is for
- + Founders shipping AI features who need them to actually work in production
- + Teams burned by demo-grade agents that drift, hallucinate, or break under real load
- + Operators wanting to add an internal-tool agent to a workflow
Who this isn't for
- − Pure research projects without a production target
- − Anyone expecting AGI / autonomous magic
- − Teams that want a chatbot wrapper, not a real working agent
D / Ownership
What you own
I don't host. I don't rent. I don't babysit. You own everything.
The code
Your repo, your IP, from day 1. No transfer drama at the end.
The infrastructure
Your AWS, Vercel, Supabase. You hold the accounts, you control them.
The AI keys & tokens
Your Anthropic / OpenAI accounts. Your spend, your limits, your control.
The data
Your database, your users. I have no access after handover.
How we work together
Free 30-min call
Scope your project with me before you commit anything. No pitch deck, no card on file.
30% kicks off
If we both want to move forward, 30% commits the project. The rest is milestone-based after each delivery.
Fixed price, honestly
Scope changes are quoted separately, never absorbed silently.
E / FAQ
Frequently asked
Which models do you support? +
Claude, GPT-4/5, Gemini, open-source. Provider-agnostic, picked per use case.
What's an eval suite? +
Tests for AI behavior. Catches regressions before production. The thing 95% of agent projects skip.
Do you handle vector DBs / RAG? +
Yes. Pinecone, pgvector, Supabase, your choice. RAG done right, not duct-taped.
What about cost monitoring? +
Built-in. You'll see token spend per request, per tenant, per feature.
Can I bring my own API keys? +
Yes. The agent runs in your environment, on your accounts.
How do you prevent prompt injection? +
Layered defenses: input sanitization, tool-permission scoping, output validation. Standard, not bolted-on.
Already prototyped?
Not ready to book a call yet? Get the 6-axis fragility model run on your URL, in a 5-minute Loom.
Get a free 5-min Loom audit →
Free · in 72h
F / Next
Ready to ship?
€25.000 · 4 weeks · Fixed price
Start your 30-min call
No commitment. No pitch deck.
What happens on the call
- 01 We talk through your build, your goals, your constraints.
- 02 We shape the smallest scope we can ship production-grade.
- 03 You walk away with scope, timeline, and a starting price.