★ AI agent · cost estimate

How Much Does It Cost to Build an AI Agent?

An agent is a chatbot that can change things: send the email, update the CRM, move the money. That makes guardrails, audit logs and human approval part of the product, not extras. Most of the hours go there.

Runs in your browser
No signup

1What are you building on?

Platform

Responsive web app

Design

A polished component library, lightly themed.

2Who builds it?

Team

One person, low overhead, slower calendar time

Blended hourly rate

$/ hour

A starting point. Type the rate from your actual quote.

3Pick the features

Your estimate: $20,100–$42,700.

Want a second opinion? Send it to me and I'll tell you which features I'd cut to ship weeks sooner.

Send me this estimate

What people forget to budget for

Human in the loop

Approval queues for risky actions, with enough context for a person to decide quickly. This is where trust is earned.

Observability

Every step, tool call and decision logged so you can answer "why did it do that?" when a customer asks.

Failure handling

APIs time out, data is missing, the model picks the wrong tool. Retries, fallbacks and safe defaults are real engineering time.

How the estimate works

Each feature has a range of hours for a senior engineer building production-quality software. Those hours scale with the number of platforms, then the estimate adds what every real project needs on top of writing features.

build  = setup + design + Σ(feature hours × platform factor)
hours  = (build + 20% QA + PM overhead) × 1.15 contingency
cost   = hours × hourly rate
weeks  = hours ÷ productive hours per week

Ranges, not single numbers, because the honest answer depends on details nobody knows yet: how clean the third-party APIs are, how many design rounds you want, how often scope moves. If someone quotes a single number without asking those questions, the range is hiding in their change requests.

Building for web and mobile is not twice the work with React Native: the backend, business logic and much of the UI code are shared. That is why the platform factor for web plus mobile is about 1.5, not 2.

Worked example

Take an AI agent MVP with 6 features (sign-up & login, admin panel, email & push notifications, dashboards & charts, third-party integrations (×2), ai agent with tools), built by a senior freelancer at $60/hour:

Worked example
Line	Hours
Feature build (incl. setup and design)	224–476
QA, project management, contingency	111–236
Total	335–712

That comes to $20,100–$42,700 over roughly 11–24 weeks. Deferring admin panel alone would save about $4,310.

Hours by feature

The ranges the estimator uses, before platform factor and overhead. The features founders most often underestimate are payments, permissions and admin tooling: none of them are visible in a demo, and all of them are needed the week after launch.

Hours by feature
Feature	Covers	Hours
Sign-up & login	Email + Google/GitHub OAuth, password reset, sessions	16–32
Admin panel	Manage users, content and orders without touching the DB	32–64
Email & push notifications	Transactional email, push, in-app, preferences	16–40
Dashboards & charts	User-facing analytics, aggregates, exports	24–56
Third-party integrations (×2)	e.g. Slack, Google Calendar, HubSpot, Zapier	24–60
AI agent with tools	Multi-step agent that takes actions, with guardrails and logging	80–160

Questions founders actually ask

How is an AI agent different from a chatbot?

A chatbot answers; an agent acts. Once software can take actions, you need permissions, approval steps and audit logs, which is why agents cost more.

Which actions should an agent take on its own?

Start with read-only and easily reversible actions. Require approval for anything involving money, customers or deleting data.

How do you test an AI agent?

With a fixed set of scenarios and expected outcomes, scored on every change, plus a sample of real runs reviewed by a person.

How accurate is this estimate?

For a well-scoped MVP, expect the real number to land inside the range, assuming scope stays put. The bigger risk is scope creep, not the estimate: every feature added mid-build costs more than it would have up front.

Why is this higher than the agency quotes I've seen?

Teaser quotes usually exclude QA, project management, revisions and a buffer for the unknown. This estimator includes all four, so it is closer to what you will actually pay.

Do AI coding tools make an MVP cheaper?

They speed up boilerplate, tests and UI scaffolding. They do not remove product decisions, integrations, edge cases or debugging in production, which is where most of the hours in this breakdown sit.

Estimate by product type

Or start from scratch with the full MVP cost estimator.

More from the workbench

02 · AI & LLMsLLM API Cost CalculatorYour monthly bill across Claude, GPT, Gemini and open models, side by side.03 · Money & fundraisingRunway & Burn CalculatorMonths of cash left, the date it hits zero, and whether you are default alive.04 · Money & fundraisingSaaS Metrics CalculatorsLTV, CAC, churn, NRR, burn multiple and more, each with the formula and benchmarks.

Fine print

Tools: on the house.
Engineer: by the hour.

Building it isn't free either, but it can be faster than you think. Send me the breakdown and I'll tell you what I'd ship first.

I'm Himanshu, an AI and full-stack engineer. I build MVPs, AI agents and the growth plumbing around them for early-stage startups.

Hire me →My projects