Keel Agents
Send a brief
Menu

AI agents, RAG assistants and LLM hardening audits for small teams

I build three things: agents that do a task and wait for a person to approve it, assistants that answer from your own documents, and a security review of an AI feature you already run. Each ends with something you own and numbers that show whether it works.

The evidence on these pages comes from two projects I built myself, to show how I work. They are not client engagements.

Send a brief See pricing

Agent workflow build
From €7,500
RAG assistant
From €4,000
LLM hardening audit
From €1,500
Retainer
From €1,800 a month

three services

Pick by the problem you have

Agent workflow build

From €7,500

Typically 6 to 10 weeks

Choose this when people copy information between tools, or watch for changes, and a person still has to check the result.

I build AI agents that read a request, draft the work and write to tools such as Asana and HubSpot, but only after a person approves. Dates and totals come from plain code; spend is capped in code.

Agent workflow build in detail

RAG assistant

From €4,000

Typically 3 to 6 weeks

Choose this when the answers already exist in your documents and people lose time finding them.

I build an assistant that answers from your own files, cites the passage each sentence comes from, and refuses when the documents do not say. It is scored against your real questions before and after every change.

RAG assistant in detail

LLM hardening audit

From €1,500

Typically 1 to 2 weeks

Choose this when an LLM feature already exists, or is about to launch, and you are not sure it is safe, affordable or any good.

I review it for prompt injection, runaway spend, missing logs and unmeasured quality, and hand you a written report with the findings ranked by risk, plus an evaluation baseline.

LLM hardening audit in detail

choosing

Not sure which one you need?

  • The work ends with something written into another system. A task, a CRM record or an email. That is an agent.

    Agent workflow build

  • The answer already exists in a document. A handbook, a contract or a knowledge base. That is a RAG assistant.

    RAG assistant

  • You already have an LLM feature, or a prototype, and you doubt it. That is the audit, priced straight from your brief.

    LLM hardening audit

  • More than one applies. Start with the smaller piece. The question set, the spend caps and the logging built for it carry over to the next.

  • None of them applies. Send the brief anyway. Sometimes the right answer is a spreadsheet formula or a no-code automation, and I will tell you before you pay for anything.

included

What every build includes

  • Scoping first. A fixed €500, invoiced up front and deducted from the build price if you go ahead. If you stop there, you keep the written scope.
  • You own the code. It lives in a repository you own, with tests, evals, documentation and a handover session.
  • Fixes for 30 days. I fix defects in what I built for 30 days after handover at no charge.
  • Spend is capped in code. Every model call is capped in code. Model usage is billed to your own provider account, with no mark-up.
  • The audit is different. It delivers a report instead. What each includes.
  • Already running something? A retainer from €1,800 a month covers monitoring, evals and improvements. See the full price schedule.

Tell me what is eating your team’s week.

Send a short brief: what the work is, who does it now, and what it costs you. I reply within two working days with a rough scope, a price range and whether AI is the right tool for it at all.

Send a brief

Or write to [email protected].

  • Defects fixed at no charge for 30 days after a build’s handover
  • Stop partway: pay for the work done, keep the code
  • No mark-up on model usage