Skip to content
MakerGain
← Back to project

An Upwork agent that stops you from bidding more, and charges when someone replies

Edited by MakerGain

Built by Ivan's product studio, Lancer watches Upwork's job feed for freelancers, filters by budget and client quality, generates and submits proposals, and collects replies in an inbox. Users remain responsible for reviewing proposals, replying to clients, and maintaining Connects. Loosening filters to bid more requires an explicit opt-in. Pay-Per-Lead charges for conversations started when someone replies, plus a base fee; Unlimited has a fixed fee and unlimited conversations. The terms do not guarantee results and state that fees are nonrefundable.

Lancer's server rejects a particular kind of change: if you try to remove the minimum budget or drop client-quality filters to let more jobs through, it blocks the change unless you explicitly opt in. The company says the rule applies equally to manual changes by people and changes made by AI agents.[1]

A company selling bidding software makes “bid a little more” a threshold you must deliberately cross. That is worth pausing over.

Lancer is an Upwork AI agent: it watches the job feed according to your criteria, assesses whether a job suits you, generates proposals from a knowledge base you write, submits them, boosts bids when worthwhile, and gathers replies in an inbox.Lancer - Upwork AI Agent[1] It is operated by MVP Masters OÜ, registered in Estonia. Its website says more than 400 freelancers and agencies currently use it for development, design, marketing, recruiting, and data work.[1]

Which part of the job does it want to take over? Its About page describes the freelancer's predicament this way: a good listing accumulates proposals within an hour, while someone asleep, on a call, or halfway through writing never sees it.Lancer - Upwork AI Agent[1] According to the company, the usual responses—refreshing the feed all day, hiring a virtual assistant to bid, or cobbling together workflow tools and spreadsheets—merely exchange one headache for another.Lancer - Upwork AI Agent[1] That is the seller's description, not a third-party measurement.

Where a person fits into the workflow

Imagine a website-redesign job with a $500 budget, a client who has spent twenty thousand dollars, and a 100% hire rate—provided your minimum budget is no higher than $500 and your client-quality criteria let it through. I made up this example to illustrate the workflow; it is not a real client.

It goes through an initial filter. The website says this reads the live Upwork job feed, with criteria including budget, client spending history, hire rate, reviews, category, and location. A second AI suitability check reads the listing in a more human-like way, specifically catching jobs that meet the filters but are clearly ones you should not take. Only the remaining jobs get proposals generated from what you put into the knowledge base—positioning, work samples, and pricing. The agent also answers the employer's screening questions, then submits, boosts when worthwhile, and tracks replies in the inbox.[1]

Image: From the product website[2], obtained on 2026-11-18; cropped from the official product interface. It is not independent evidence of effectiveness.

People are not excluded. The company says proposals can be reviewed before sending, campaigns can be paused, and each AI decision can be inspected and adjusted. The homepage also lists a human-approval mode, automated-bidding delays, and Discord/Slack notifications, including alerts when Connects run out or a bid fails.[1] · Lancer - Upwork AI Agent[2] The terms are blunter: Lancer is software you operate, not a recruiting, staffing, or lead-generation agency. It does not manage your Upwork account or sales. You must review campaigns, the knowledge base, and cover letters, reply to clients yourself, and buy and maintain Connects—Upwork's bidding credits—directly with Upwork.[3]

The maker, and the line he drew for himself

The person behind the product is Ivan Nedelkovski. He ran a software-development consultancy, sold it in 2025, and opened a product studio. Lancer was its first product. The interview says the idea passed his own validation criteria before he started building: he was scratching his own itch.[4]

One product choice does fit that account. The website presents having “most functions operated by an AI agent rather than a person” as a design choice. That is why the guardrail has to govern agents too.[1]

The rule that appears to work against revenue

Back to that opening guardrail. For a company selling bidding software, stopping users from bidding more would normally work against revenue—unless it does not charge by the number of proposals.

The pricing page turns that into a slogan: others charge per proposal; we charge per “conversation started,” meaning it counts when someone replies. But PPL is not purely results-based: there is a fixed base fee, with additional charges beyond the included allowance.Lancer - Upwork AI Agent[5] With quarterly billing, Pay-Per-Lead is listed at $99/month, including 15 leads per quarter, then $19 per additional lead. Monthly billing is $149/month with 5 leads per month. Unlimited is listed at $333/month with quarterly billing and unlimited conversations, or $499 paid monthly. Both plans include all features. PPL includes one Upwork account; Unlimited includes multiple accounts, Done-For-You setup, one hour-long strategy consultation, and monthly follow-ups.[5]

On revenue, the interview body says Lancer reached $10k MRR two months after launch and lists “current” MRR of $20K with a two-person team. Yet the Revenue field at the top of the same page says “>$25K a month.” The $10k refers to the two-month milestone; $20K is the body's “current” figure. The header's “>$25K” does not match that $20K, and the page does not display the interview's publication date. I am not choosing one for it.[4]

One point needs to be clear: MRR is not profit. Public information gives no infrastructure, AI-token, payment-processing, or staffing costs, so whether this business actually makes a profit cannot currently be calculated.

Its candor is also where the buyer takes the risk

One passage in the terms is uncomfortable but candid: subscriptions are charged automatically through Stripe at the start of each billing period, along with usage-based charges, until the user cancels. All fees are nonrefundable, including partial billing periods, unused proposals/leads/credits, quarterly and annual plans, and one-time setup or onboarding that has already begun. Results falling short of expectations do not qualify for a refund either.[3]

In terms a buyer can feel: you are buying a machine for starting conversations. It cannot guarantee those conversations become contracts, and the money does not come back if they do not. The contract explicitly says it does not guarantee any number of proposals, replies, interviews, hires, contracts, or revenue.Lancer - Upwork AI Agent[3] It is not an appealing clause, but at least it does not pretend to guarantee results.

Lancer's website says Upwork prohibits bots that spray out bids—and says that prohibition is right. It presents Lancer as moving in the opposite direction: fewer proposals, each written for the specific job.Lancer - Upwork AI Agent[1] This is Lancer's account, not an endorsement by Upwork, nor does it mean Upwork approves of such tools.

The contact page has a revealing detail: sales demos use your own Upwork profile and the jobs you want to bid on, not a preset demo account.Lancer - Upwork AI Agent[6] That at least turns “can it bid in my obscure specialty?” from an abstraction into a sample you can see on the spot, although a demo itself cannot prove bidding accuracy. The homepage also carries a customer account: Inboxist's founder says Ivan approached him with a free trial while he was still subscribed to a competitor. He switched after trying it, despite having two months left on the competing subscription.Lancer - Upwork AI Agent[2] This is a testimonial on the seller's website, not independently verified. Advertising, trial-conversion, and CAC data are not public either.

What this business really sells is not more proposals, but the moment someone replies. Whether that moment leads to an order or more conversations that go nowhere rests on the paying customer's own profile, pricing, and response speed. It gets at least one thing right: when you want to loosen the standards and let the machine submit a few more bids, it stops you first.

Continue reading

AI productivity
Klipy, the sales AI that waits for your approval

Klipy takes over follow-up after sales meetings. It captures customer conversations from email, meetings, and instant messaging, drafts follow-ups and CRM updates, and prepares for the next meeting. Follow-up messages go out only after the salesperson approves them. Founder Jung Hong Kim is a Korean serial entrepreneur who took on debt when COVID sank his retail venture. He met his cofounder while consulting and launched Klipy.ai in three months. It went live in November 2024 as an AI CRM that logged communications automatically. In June 2025, it shifted toward carrying out next steps for sales teams and tied billing to specific actions users approve.

AI productivity
The founder reported \$8,000 in 2024 revenue, 85% collected only once: NodeLand turns long reports into maps

Rio de Janeiro-based Gabriel Xará runs NodeLand as a side project. Canvas extracts material from long reports, video transcripts, or meeting notes and generates structured maps; people still decide which nodes to keep, whether relationships are correct, and who the map is for. Built in 2023 and launched in 2024, it acquired early users through platforms such as AppSumo. A January 2025 interview disclosed \$8,000 in 2024 revenue, 85% from lifetime licenses, with costs and profitability undisclosed. He had focused almost entirely on the product that year, with limited marketing.

AI productivity
After closing a product nobody wanted, they broke entrepreneurship into “the next step”

David Heikka and his brother Felix launched their first AI form product and found zero demand. They tried ads, landing-page changes, renaming, and added features, then closed it after six months. They went on to build Buildpad, breaking business creation into stages that indicate what to validate next. Its early and current billing mechanisms differ, and the product was renamed aicofounder before changing back to Buildpad.

AI productivity
The client owns the code and pays the AI bill: SpaceAI360 builds automation you can take with you

Founded by Shahdin Salman alone, SpaceAI360 builds portable AI workflows inside customers' own accounts. It handles parsing, OCR, field extraction, database checks, and ERP/CRM writes after a form, email, or file arrives, routing uncertain items to human review. Rather than generic AI consulting, he starts with the highest-friction bottlenecks, assigning deterministic steps to n8n and judgment steps to Claude.