Free tool
AI app development cost estimator
What an AI app, agent or AI feature costs to build, and what it costs to run each month once people use it. Build ranges are the prices I publish for this work; running costs use current OpenAI, Claude and Gemini rates.
To build
$15,000 to $40,000
8-14 weeks. With your scope, expect the lower end of the range.
To run, a month
$390
On a mid-priced model such as Claude Sonnet 5.5, for 30,000 model calls. From $19.50 on GPT-6 Luna to $1,950 on Claude Fable 5.1. API only: hosting, storage and embeddings are extra.
First year
$19,680 to $44,680
The build plus twelve months on the mid-priced model.
Ranges are my own published prices for AI product build. Email the estimate and I reply with a fixed quote once the scope is clear.
Where the build figures come from
Each kind of work is one of my offers, with the length and typical budget its page publishes:
- A new AI product, from an empty repo: AI product build, 8-14 weeks, typically $15k-$40k.
- An AI feature in a product you already run: RAG development, 2-5 weeks, typically $4k-$10k.
- A prototype that works in a demo and has to reach production: AI production rescue, 2 weeks to stabilise, then ongoing, typically $3k-$6k.
Separate customers, billing, integrations, compliance and real-time features each add work, so the more of them apply, the higher in its range a build lands. My own multi-tenant AI support platform, Evoriqa, went from first commit to live in about two months.
How the running cost is worked out
Users times actions per user gives the number of actions a month. Each action is one or more model calls of a typical size:
- Short chat replies: one call of about 1,500 tokens in and 300 out.
- Answers from your documents (RAG): one call of about 4,000 tokens in and 500 out.
- Agent tasks (about 8 model calls each): 8 calls of about 3,000 tokens in and 400 out.
For exact token counts and every model side by side, use the API pricing calculator. For what drives each figure up, read what it costs to build an AI app. Not sure you are ready to build at all? Take the AI readiness assessment first.
Questions
- How much does it cost to build an AI app?
- A new AI product built from an empty repo to production typically costs $15k-$40k over 8-14 weeks in my own work. Adding one AI feature, such as answers from your documents or an agent, to a product you already run is typically $4k-$10k over 2-5 weeks. Scope moves the figure within each range.
- How much does an AI app cost to run each month?
- The model API bill depends on how many model calls the app makes and how many tokens each carries. A thousand users each asking thirty questions a month of their documents comes to about $20 on GPT-6 Luna, $390 on Claude Sonnet 5.5 and $1,950 on Claude Fable 5.1. Hosting, storage and embeddings come on top.
- Why does an AI agent cost more to run than a chatbot?
- An agent calls the model several times for one task, to plan, act and check, and each call resends the context. Eight calls per task cost roughly eight times one chat reply, before any retries.
- Are these figures a quote?
- No. They are the typical ranges I publish for each kind of work and current API list prices. A fixed price comes in writing once the scope is agreed.
Tell me what you’re building and where it’s stuck.
I’ll tell you the cleanest path forward, including if it’s “don’t build that.”
Email only, on purpose: it reaches me, not a sales team.
Or write tocontact@alihassan.dev
