Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Everyday tasks

Set a freelance rate that survives fees, bidding costs, and review time

Written October 26, 2026 · Edited November 8, 2026

For a freelancer × setting rates × an agent doing the arithmetic, ending with a rate you can defend in a proposal: a target take-home figure, divided by what you actually keep, plus the real costs of winning and doing the work — including the time you spend checking the agent's output.

Before you start

You need last month's real numbers, or at least honest estimates for them: what you earned, what you spent winning work, and what your tools cost. The agent's job here is arithmetic and bookkeeping, not judgment: it can divide, total, and compare, but only you can say what your time is worth and how much unpaid work you are willing to carry. The platform figures below are Upwork's own published pricing, because that is where most readers of this page will bid — but the method works on any platform once you swap in that platform's fee. No rate on this page is a recommendation for your work.

The one discipline the whole workflow rests on: the rate is set from the take-home figure backward, never from the headline figure forward. A rate that looks right before fees and costs is not a rate; it is a hope.

What "done" looks like

At the end of one sitting you have a one-page rate sheet: your target take-home per hour, the fee on your next contract, your bidding cost per client won, your monthly tool cost spread across billable hours, and your review time per hour of agent-assisted work — combined into one hourly rate and one fixed-project formula, with a worked example you can re-run when any input changes.

Steps

  1. Pick the take-home figure first. Write down what you need to keep per billable hour after every deduction — not what you want to charge, what you kept. This is the only number in the workflow that is a judgment call, so make it deliberately: it has to cover your living costs, taxes, unpaid time, and time off, spread across the hours you actually bill. If you bill 20 hours in a typical week, a target that assumes 40 billable hours is not a target; it is a shortfall with a plan attached. The agent can help you spread annual costs across billable hours, but it cannot decide what your time is worth.
  2. Divide by what you actually keep. Upwork's pricing page says freelancers pay a variable service fee ranging from 0% to 15% on earnings, shown before you submit a proposal — so the fee on your contract is a known number, not a guess. The arithmetic is simple: rate equals take-home divided by one minus the fee. At a 10% fee, a $50 take-home needs a $55.56 rate, because $55.56 minus 10% is $50. Upwork's own example works the same way: a contract with a 10% fee and a $500 bill pays $450. Never use a fee percentage from an article, including this page, when your proposal shows you the actual fee — the shown fee is the one that governs your contract.
  3. Add the cost of winning the work. Bidding is not free, and a rate that ignores it quietly subsidizes every lost bid with the won ones. Total last month's bidding spend — Connects bought, at $0.15 each on Upwork's pricing page, plus any membership cost for the month — and divide it by the clients won that month. That is your acquisition cost per client. Spread it across the expected hours of a typical contract from that client: if winning a client cost $30 and the first contract runs 10 hours, add $3 per hour to the rate. If you won no clients last month, the honest input is not zero; it is last month's spend with no return yet, which is a reason to fix your bidding before raising your rate.
  4. Add the tool cost, spread honestly. List every tool the work actually depends on: the agent subscription, the design or editing tool, the invoicing tool, and anything else you pay for monthly to do the job. Total the monthly cost and divide by your monthly billable hours. A $20 monthly tool spread across 80 billable hours is $0.25 per hour; the same tool spread across 20 billable hours is $1.00 per hour. The point of this step is not precision to the cent; it is refusing to pretend tools are free. A tool you pay for but never use is not a cost to spread — it is a subscription to cancel.
  5. Add the review time the agent still costs you. This is the step most agent-assisted rate math skips, and it is the one that matters most once agents do the drafting. Time three recent agent-assisted jobs: the agent's drafting time, and your time checking, correcting, and reworking the output before it was fit to send. If an hour of agent drafting needs 20 minutes of your review, then every hour of agent output costs you a third of an hour of your own time, and that time is billable time spent, whether the client sees it or not. Add that fraction to the hours you estimate for the job, not to the rate directly: a job the agent drafts in 2 hours with 40 minutes of your review is a 2-hour-40-minute job, priced at your rate, not a 2-hour job with a discount for the agent's help.
  6. Combine the inputs into one rate, and one project formula. Hourly rate: take-home divided by one minus the fee, plus acquisition cost per hour, plus tool cost per hour. Fixed project price: estimated hours including review time, multiplied by that hourly rate — the fee is already inside the rate, so do not divide by it a second time. Run one worked example end to end with your own numbers before using the rate anywhere: a formula you have not run with real inputs is a guess with algebra on top.

    Combine the inputs into one rate. Take the numbers from the steps above: a $50 take-home at a 10% fee needs a $55.56 base rate ($50 ÷ 0.90); add the $3.00 an hour acquisition cost from Step 3 and the $0.25 an hour tool cost from Step 4, and the rate is $58.81 an hour — one addition line: $55.56 + $3.00 + $0.25 = $58.81.

    For a fixed project, multiply the hours including review time by that rate: the 2-hour draft plus 40 minutes of review from Step 5 is 2 hours 40 minutes, so 2⅔ × $58.81 = $156.83. The fee is already inside the rate, so do not divide by it a second time.

    Check it by reversing the steps: $58.81 minus the $3.00 acquisition and $0.25 tool costs leaves $55.56; keeping 90% of that is $55.56 × 0.90 = $50.00 — the take-home figure you started with.

  7. Compare the result against the market you actually bid in. If your computed rate is far above what comparable postings pay, the answer is usually not to cut the rate blindly — it is to check which input is doing the damage. A high acquisition cost means your bidding needs work, not your rate needs cutting. A high tool cost means a subscription needs cutting. A high review fraction means the agent workflow needs tighter prompts or a narrower task, not cheaper hours. Cut the input that is actually wasteful; cutting the rate to win work you lose money on is not a strategy, it is a slow exit.
  8. Re-run the sheet monthly, and keep the old sheets. Fees change per contract, bidding costs move with your win rate, tool prices change, and your review fraction falls as your prompts improve. Re-run the sheet at the start of each month with last month's real numbers, and keep each month's sheet. The history is what turns a rate from a feeling into a record: when a client asks why your rate moved, the sheet is the answer, and when your own costs creep up, the sheet shows it before your bank balance does.

Check the agent's work

What can go wrong

Limits

Related playbooks

Sources

Sources: Upwork freelancer pricing.

Back to: Everyday playbooks hub · Also: Write Upwork proposals with an agent, without inventing anything