Eight providers read directly
Hire prompt engineers and prompt engineering consultants by the hour
You hire a prompt engineer when an LLM feature works in the demo and wobbles in production: answers drift, the format breaks, the token bill climbs and nobody can prove a prompt change helped. We read the hire pages of eight providers, turned every salary band into an hourly cost, and worked out when a contract prompt engineer costs less than a full-time hire.
Provider pages read directly · No paid placements · Scope the prompt work in the form
Hire brief generator
demo · one brief per browser
That was the one free brief for this browser.
To hire for the job, put it on the public BotGigs job board: $99 a month on Growth, applications straight to your inbox, cancel any time.
Post a job, $99 a monthfor a different task
Best-effort estimate, not a quote or a match
We could not scope your task just now, so this is an example brief for a different automation, shown so you can see the format. Your free brief has not been used: try again in a minute.
scope of work
who to hire
Screen for
Effort estimate
questions to ask your hire
- ?
Like the brief? Put the job in front of specialists on the board.
the short answer
To hire a prompt engineer on contract, expect $60 to $100 an hour on Arc.dev and $80 to $250 an hour on MentorCruise, with $140 as its typical rate. A full-time US prompt engineer earns $97,940 (ZipRecruiter median) to $135,000 (KORE1 placement median), which is about $16,070 a month fully loaded. Toptal, LeewayHertz and Width.ai publish no rate at all. Most teams do not need a full-time role: an audit of one production feature with a proper test set is about 30 to 60 hours, or $3,000 to $8,400 at $100 to $140 an hour, which is less than three weeks of a full-time hire.
The audit
What prompt engineering services publish on their own hire pages
Eight providers that show up when a US company searches for a prompt engineer, each read from its own page in September 2026. Where a figure is a salary rather than a bill rate, the row says so, because the two are not the same number.
| Provider | Model | Published price | Time to start | Where the talent is |
|---|---|---|---|---|
| Arc.dev | Freelance or full-time marketplace, no fee until you hire | $60 to $100+ an hour | Freelance within 72 hours, full-time within 14 days | Global pool across 190 countries |
| MentorCruise | Expert sessions and monthly plans, 7 day refund trial | $80 low, $140 typical, $250+ senior an hour. Monthly from about $120 | 48 hours | Mixed |
| Tecla | Placement, 90 day replacement guarantee | Salary bands. US: $100k to $140k junior, $140k to $190k mid, $190k to $260k+ senior. Latin America: $45k to $115k | 5 day average placement, 2 to 3 weeks to onboard | US and Latin America |
| KORE1 | Staffing firm: direct hire, contract (3 to 6 months), contract to hire | Salary: $102k to $165k mid-level, $135k placement median | 25 to 40 days to fill | US |
| Toptal | Hourly, part-time or full-time, trial of up to two weeks | Not published | Match in under 24 hours, hire in about 48 | Worldwide |
| Width.ai | On-staff engineers, billed monthly, no long commitment | An hourly rate is mentioned, no figure | 48 hours, same day start possible | US and overseas |
| LeewayHertz | Dedicated team, team extension or fixed project | Not published | Not stated | Not stated |
| Turing | Only a job page for engineers seeking prompt work | No company hire page | Not applicable | Not applicable |
Two of eight publish a bill rate you can budget against, two publish salary bands that tell you what a hire costs rather than what an hour costs, and four publish nothing usable. That split matters because prompt work is usually a project, and the providers that hide the rate are also the ones selling dedicated monthly engineers. Ask every provider the same three things in writing: the hourly rate, the minimum commitment, and whether the engineer who interviews is the one who does the work. Independent salary trackers land in the same range as the staffing firms: ZipRecruiter lists a $97,940 median for "Prompt Engineer" (about $47 an hour), and Glassdoor shows a $131,483 average, though Glassdoor's figure rests on only 32 salaries.
The unit that matters
Prompt engineer salary bands turned into an hourly cost
A salary and a contract rate only compare once they are in the same unit. We divided each published salary by 2,080 working hours. Wages and salaries are about 70 percent of what a US private employer spends on a worker, according to the Bureau of Labor Statistics compensation survey, so the loaded column divides once more by 0.70 to add benefits and payroll taxes.
| Published figure | Salary | Per hour, wage only | Per hour, fully loaded |
|---|---|---|---|
| Tecla, Latin America mid-level | $60,000 to $85,000 | $28.85 to $40.87 | $41.21 to $58.38 |
| ZipRecruiter median, Prompt Engineer | $97,940 | $47.09 | $67.27 |
| Glassdoor average (32 salaries) | $131,483 | $63.21 | $90.30 |
| KORE1 placement median | $135,000 | $64.90 | $92.72 |
| Tecla, US mid-level | $140,000 to $190,000 | $67.31 to $91.35 | $96.15 to $130.49 |
| Tecla, US senior | $190,000 to $260,000 | $91.35 to $125.00 | $130.49 to $178.57 |
Read the table against the contract rates above and one thing jumps out. The KORE1 median prompt engineer costs an employer about $92.72 an hour before any recruiting fee, idle time or management overhead, which is the same number as the top of Arc.dev's $60 to $100 band. So a contractor at $100 is not a premium over employment; it is roughly the same hour without the 65 to 90 day search. The same arithmetic works as a floor. The Bureau of Labor Statistics puts the median US software developer at $135,980 a year, and a firm that pays its engineer 50 to 70 percent of the bill rate has to charge about $93 to $130 an hour to employ someone at the KORE1 median. Rates well under that are either from outside the US or from someone early in the career, which may be fine, but you should know which one you are buying.
Contract or full-time
When a freelance prompt engineer stops being cheaper than a hire
A full-time prompt engineer at the $135,000 KORE1 median costs about $192,900 a year loaded, or $16,070 a month, whether the month has 20 hours of prompt work in it or 200. Divide that by a contract rate and you get the number of hours a month at which the two cost the same.
| Contract rate | Source | Hours a month to match a $16,070 hire | Verdict |
|---|---|---|---|
| $60 an hour | Arc.dev, bottom of band | 268 | Contract always wins; a full month is about 173 hours |
| $80 an hour | MentorCruise, low end | 201 | Contract always wins |
| $100 an hour | Arc.dev, top of band | 161 | Hire only if the work is nearly full time for a year |
| $140 an hour | MentorCruise, typical | 115 | Hire if you need two thirds of a person, every month |
| $250 an hour | MentorCruise, senior | 64 | Use for audits and reviews, not ongoing work |
Very few companies have 115 to 160 hours of prompt work every month for a year. Prompt work comes in bursts: a new feature, a model upgrade that changes behavior, a cost spike, a customer complaint that exposes a failure mode. Between bursts the job shrinks to a few hours of monitoring. That is why the honest answer for most teams is a scoped contract now and a small monthly retainer after, and why a full-time prompt engineer usually ends up doing broader generative AI development within six months anyway.
What you are paying for
What prompt engineering services should deliver, with typical hours
A prompt engineer is worth hiring for the measurement, not the wording. These are the scopes we would put in a hire brief, with effort bands and what they cost at $100 to $140 an hour.
Prompt audit of one production feature
30 to 60 hours, $3,000 to $8,400
Read the prompts, collect 100 to 200 real inputs with expected outputs, score the current version, rewrite, and score again. You get a number, not an opinion.
System prompts and tool calling for an agent
60 to 120 hours, $6,000 to $16,800
Instructions, few-shot examples and tool descriptions for a support or operations agent, tested on your own tickets, with the failure cases written down.
Token cost reduction pass
20 to 50 hours, $2,000 to $7,000
Shorter prompts, prompt caching, routing easy requests to a cheaper model and structured outputs that stop retries. Priced against your current monthly model bill.
Model migration check
15 to 40 hours, $1,500 to $5,600
When a provider retires a model or you switch vendors, the same prompts behave differently. Run the test set on both, fix what broke, sign off before the cutover.
Retrieval prompt tuning for RAG
30 to 80 hours, $3,000 to $11,200
Query rewriting, citation formats and refusal rules for an assistant over your documents, tested against questions your users actually ask.
Monthly prompt retainer
10 to 20 hours a month, $1,000 to $2,800
Watch the scores, review failures users report, adjust prompts and keep the test set current. The part most teams forget until quality slides.
Every one of those scopes starts with the same artifact: a test set built from your real inputs. It is the thing that separates prompt engineering from rewording, and it outlives the engagement. Our guide to evaluating an AI agent before you ship it shows how to build one, and if the feature answers from your documents, check whether retrieval is working before paying anyone to tune the prompt, because a better prompt cannot fix the wrong document.
Before you hire
Do you need a prompt engineering consultant, or a different hire?
Hire a prompt engineer when the model is right, the data is right and the output still is not: inconsistent tone, broken JSON, answers that ignore instructions, or a token bill that grew faster than usage. Those are prompt and measurement problems, and a specialist fixes them in weeks.
Hire an AI engineer instead when the problem is the system around the prompt. If the assistant retrieves the wrong documents, times out, cannot call your CRM or has no logging, rewriting instructions will not help. That is integration and infrastructure work, and it belongs with AI engineers or an LLM development company that owns the whole build.
Hire nobody yet when you cannot write down what a good answer looks like. Collect 30 examples of bad outputs and what they should have said. That list is the brief, and it is also the first test a candidate should pass. If you are still pricing the build itself, our AI agent development cost audit breaks down where the hours go.
How it works
From a flaky LLM feature to a prompt engineer with a score to beat
Describe the feature and what goes wrong
In the hire brief at the top of this page, name the model and provider, what the feature does, and the failures you see. Two minutes.
Get a scoped brief with hours
The brief comes back with deliverables, an effort band in hours and the skills to screen for, so you can compare it with the rates above in the same unit.
Pick a prompt engineer who has shipped on your model
Matched specialists show their rate and the production LLM work they have done. You agree the rate directly, with no hire fee on the Growth plan.
Start with a paid trial on 30 real failures
Ask for a scored before and after on your own inputs, plus token cost per request. Extend the engagement only when the score moves.
The honest version
Where BotGigs fits, and where another route is better
BotGigs is a marketplace for AI and automation specialists. Buyers pay $99 a month on the Growth plan to keep a job on the board, with no hire fee, and agree the rate with the specialist directly. See BotGigs pricing for the plans. Every specialist is an AI or automation builder, so you are not filtering prompt work out of a general freelance crowd.
If you want a two week trial with a replacement guarantee and do not mind an undisclosed rate, Toptal fits. If you are filling a permanent US seat and want a recruiter to run the search, a staffing firm such as KORE1 or Tecla is built for that. If you need one hour of advice rather than a deliverable, an expert session on MentorCruise is cheaper than any engagement. For a scoped prompt audit, a cost pass or an agent's instructions written and measured by someone who has done it in production, post the job here.
Questions
Questions buyers ask first when they hire a prompt engineer
How much does it cost to hire a prompt engineer?
Contract prompt engineers bill about $60 to $100 an hour on Arc.dev and $80 to $250 on MentorCruise, where $140 is typical. A full-time US hire runs $97,940 (ZipRecruiter median) to $135,000 (KORE1 placement median) in salary, which is roughly $192,900 a year once benefits and payroll costs are added. Most teams need a scoped project of 30 to 120 hours, not a full-time role.
Do I need a prompt engineer or can my developers do it?
Your developers can write the first prompts. You need a specialist when output quality is inconsistent in production, token costs are climbing, or nobody can say whether a prompt change made things better or worse. Those are measurement problems. A good prompt engineer arrives with an evaluation set and a before and after score, not just rewritten prompts.
What do prompt engineering services include?
A useful engagement includes an audit of the prompts behind one production feature, a test set of 100 to 200 real inputs with expected outputs, rewritten system prompts and few-shot examples, a cost pass on model choice, caching and prompt length, and handover documentation so your team can change prompts safely after the engineer leaves.
Should I hire a freelance prompt engineer or a full-time one?
Freelance or contract, unless prompt work fills most of a working month for a year or more. A $135,000 hire costs about $16,070 a month fully loaded. At $100 an hour a contractor only costs more past about 161 hours a month, and at $140 past about 115 hours, while a full-time search takes 65 to 90 days in house.
How long does it take to hire a prompt engineer?
Through a talent marketplace, one to five days: Toptal quotes a match in under 24 hours, Arc.dev within 72 hours for freelancers and Tecla a five day average placement. A full-time hire through a staffing firm takes 25 to 40 days according to KORE1, and 65 to 90 days when the company recruits in house.
What is the difference between a prompt engineer and an AI engineer?
A prompt engineer shapes how a model is instructed and measured: system prompts, examples, output formats and evaluation sets. An AI engineer builds the software around the model: retrieval, tool calling, APIs, deployment and monitoring. In practice the best prompt engineers write Python and the best AI engineers write good prompts, so hire on the problem you have.
How do I vet a prompt engineer?
Give a paid trial on your own feature. Hand over 30 real inputs where the current prompt fails and ask for a scored comparison of old and new prompts on those inputs, plus the token cost of each. Anyone who returns clever rewrites without a score has not done the job you are hiring for.
Get started
Hire a prompt engineer who brings a score, not just new wording
Describe the LLM feature and what goes wrong, and get back a scoped hire brief with hours and matched prompt engineers, with rates visible before you commit.
No card required to start.
Related