Estera: Product, team and how it was built
Estera (AI receptionist agent that answers phone calls and WhatsApp messages for small businesses, Romania): More than 700 users with 20 paying, August 2026. Evidence grade E: no revenue figure was checked. Team: 2 co-founders, Andrei-Fredy Crăciun and Rareş Toma, working on it since November 2025; total headcount not published. Launched February 2026; still running, revenue growing.Added 5 November 2026
- Outcome
- More than 700 users with 20 paying, August 2026Revenue growing
- Evidence
- E no revenue figure was checked
- Launch
- February 2026
- Status
- Still running
- Team
- 2 co-founders, Andrei-Fredy Crăciun and Rareş Toma, working on it since November 2025; total headcount not published
- Outside funding
- None (bootstrapped, August 2026); a round of more than EUR 1 million planned
- Main channel
- Not stated
What is not verified
No revenue figure was found. The only traction numbers are what co-founder Andrei-Fredy Crăciun told ZF on 6 August 2026 (more than 700 users, 20 paying, more than 6,400 calls and messages), and he said the team's internal target before fundraising was to pass EUR 10,000 in monthly recurring revenue, which suggests it was below that then. No TrustMRR profile or other payment-processor profile was found. The terms, privacy policy and data processing agreement are on app.tryestera.com, whose robots.txt disallows ClaudeBot, so they were not read. A reader could look for a payment-processor profile or a later interview with revenue. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.
Estera sells an AI receptionist: a program that answers a business's phone calls and chat messages in place of a person at the front desk. The phone part is a voice agent, software that turns the caller's speech into text, lets a large language model (an AI model trained on very large amounts of text) decide what to say, and speaks the answer with a synthetic voice. The WhatsApp part uses the WhatsApp Business API, the official interface through which companies send and receive WhatsApp messages from software.
Two Romanian founders, Andrei-Fredy Crăciun and Rareş Toma, started work in November 2025, after pausing their AI website builder Loopple AI. In an interview with the Romanian business daily Ziarul Financiar (ZF) published on 6 August 2026, Crăciun said the platform had more than 700 users, of whom 20 were paying, had handled more than 6,400 calls and messages, and had no outside funding. The pricing page lists plans from EUR 49 a month.
No revenue figure was found in any source this page could download, so the evidence grade is E. What is checked is that the product is live with published prices, that a national business newspaper interviewed a co-founder, and that it launched on Product Hunt on 27 July 2026.
How much revenue does Estera make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Users and paying customers, August 2026 | more than 700 users, 20 of them paying | C | Co-founder in ZF, 6 August 2026: "numărul utilizatorilor a urcat de la circa 100 la peste 700, dintre care 20 sunt plătitori" (the number of users rose from about 100 to more than 700, of whom 20 are paying) (translation). | [4] |
| Calls and messages handled, August 2026 | more than 6,400 | C | ZF, 6 August 2026: "platforma a procesat peste 6.400 de apeluri şi mesaje". The company's home page in November 2026 shows a smaller counter, "5,200+ conversations and calls handled". | [4, 1] |
| Founders | Andrei-Fredy Crăciun (product, sales, customer success) and Rareş Toma (technical development) | C | About page, which writes them "Craciun Andrei-Fredy" and "Toma Ioan-Rares". Product Hunt lists them as makers Fredy A. Craciun and Toma Rareş. | [3, 7] |
| Start of work and first version | November 2025; first version about February 2026 | C | ZF: work began "din noiembrie 2025"; the first version was launched "în urmă cu circa şase luni" (about six months before August 2026) (translation). The month February is derived from that. | [4] |
| Funding | none in August 2026; plans to raise more than EUR 1 million at a EUR 5 million pre-money valuation | C | Co-founder in ZF. The round is planned for late 2026 or early 2027, after monthly recurring revenue passes EUR 10,000, an internal target; until then they want to stay bootstrapped. | [4] |
| Earlier products | Simmmple and Loopple AI | C | ZF calls Estera the co-founder's third technology product in five years. Loopple AI, an AI website builder, is paused. The company's press release says the founders' earlier AI products reached more than 2 million users worldwide and that they were in the Forbes 30 Under 30 Romania list; not checked here. | [4, 5] |
| Prices, November 2026 | EUR 49 to EUR 279 a month, Enterprise on request | C | Pricing page. The ZF article of August 2026 gives the same range. | [2, 4] |
| AI models behind the agent | Gemini and OpenAI models with automatic fallback | C | ZF: the platform switches from one model to the other when one fails or an answer is below a quality threshold, plus in-house systems that check transcripts. | [4] |
| Product Hunt launch | 27 July 2026, ninth of the day, 137 points | C | Product Hunt product page ("# 9 Day Rank", "137 points"); launch date from Product Hunt's official API record. | [7] |
| Customers | mostly medical; also real estate, hotels and restaurants, law and consulting firms | C | ZF. Users also in Spain, France, Italy, Hungary, Switzerland and the United Kingdom, although sales target Romania. The site names Neocasa Real Estate, Remenix, SkyLine Properties and Eva Estate in case studies. | [4, 1] |
| Legal entity | Estera AI LLC | C | Copyright line on the home page. The state or country of registration is not given on the pages read. | [1] |
| Time from first version to the August 2026 figures | about 6 months | calc | Calculated: February 2026 to August 2026. | [4] |
How to read this table: a grade C value is stated by the company on its own site, in its press release, or by a co-founder in an interview. A "calc" value was computed on this page. No value comes from a payment processor, a filing or a journalist's own check.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Lite | 46.88 | 54.70 | 150 minutes a month, 1 phone number, WhatsApp not listed |
| Essential | 93.77 | 110.51 | 300 minutes, 2 phone numbers, WhatsApp and web chat included |
| PRO | 179.72 | 210.98 | 600 minutes, 2 phone numbers, CRM integration, 90+ languages |
| Premium | 264.56 | 311.45 | 900 minutes, 3 phone numbers, multi-location routing, API access |
| Enterprise | custom | custom | custom volume and contract |
Euro prices on the page: Lite EUR 49 a month or EUR 42 billed yearly, Essential EUR 99 or EUR 84, PRO EUR 189 or EUR 161, Premium EUR 279 or EUR 237; yearly billing is 15% off. The plans are priced by minutes of calls or conversations a month. Building and testing an agent is free with test credits; payment starts when the agent is published on the customer's own number or WhatsApp. When the minutes run out, the agent pauses until the customer upgrades. CRM integration on PRO has an extra setup fee. Payments run through Dodo Payments.
When did Estera launch, and how did it grow?
From the first work in November 2025 to self-serve signup and a Product Hunt launch in the summer of 2026, read from the company's pages and two Romanian press texts.
- November 2025CAndrei-Fredy Crăciun and Rareş Toma start work on Estera after pausing Loopple AI, an AI website builder that faced large funded rivals. [4]
- About February 2026CFirst version launched; each customer is set up by the team on a call. [4]
- 22 June 2026CPress release on the Wall-Street.ro release platform: Estera's agents are used by law firms, dental clinics, restaurants and hotels, and real estate agencies. [5]
- July 2026CSelf-serve setup: a customer can build a voice or WhatsApp agent in under three minutes. Users rise from about 100 to more than 700 within a month. [4]
- 27 July 2026CProduct Hunt launch; ninth product of the day with 137 points. [7]
- 6 August 2026CZF interview: 20 paying customers, more than 6,400 calls and messages handled, no outside funding, a round planned after EUR 10,000 in monthly recurring revenue. [4]
- November 2026CThe site lists five plans, 26 industry pages, four customer case studies and the outbound tool Estera Messages. [1, 2, 6]
Why did Estera work?
Four causes, in the order they mattered. Each one is a reusable lesson.
Self-serve setup
Until July 2026 the team configured each account on a call. The co-founder told ZF that local customers did not want 20-minute setup calls, and that after a three-minute self-serve setup the number of users went from about 100 to more than 700 in one month.
Usage-based plans with a free build
The pricing page lists plans by minutes of calls or conversations, not by seats. Customers build and test the agent free and pay only when they connect it to a real phone number or WhatsApp account.
A region the funded rivals ignore
The co-founder told ZF that better-funded voice-agent companies from Germany and Austria focus on Western Europe and the United States, and that Estera aims at Eastern Europe, where current models now speak Romanian well.
Two model providers with fallback
The agent switches between Gemini and OpenAI models when one fails or gives a weak answer, as the co-founder described it to ZF. This reduces the risk that one provider's outage stops a customer's phone line.
How it was built
These details come from the company's pages and the co-founder's ZF interview. They are grade C. [4, 1, 8]
- Language models from Google (Gemini) and OpenAI, with automatic switching when one returns an error or an answer below a quality threshold.
- In-house systems that check call transcripts and analyse conversations; every call is recorded and transcribed.
- A knowledge base filled from the customer's website, PDF documents or notes, which the agent updates from customer feedback.
- WhatsApp through the official WhatsApp Business API, as the company's FAQ states; integrations with Google Calendar, Shopify, WooCommerce, CRMs and webhooks.
- Customer chooses the voice, the industry and the goal of the agent, and whether it answers always, only after hours, or as backup when reception does not pick up within five seconds.
- A monthly performance report emailed on the first of each month lists questions the agent could not answer and suggests prompt fixes.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Estera, and why did they fail?
Three companies in our database of failed startups promised to answer customers or handle a busy person's communication with software or remote staff, and failed. The values below are from that database and are not independently verified.
Answerly
The promiseFAQ chatbots that deflect customer questions so people handle only the hard ones.
Why it diedThe language technology of the time could not understand varied questions, and Zendesk and Intercom already owned the support desk.
Vanti
The promiseAI chatbots that hold sales conversations with a brand's customers.
Why it diedNo product-market fit in a crowded conversational AI market; high customer acquisition cost and hard integrations.
Zirtual
The promiseRemote human assistants who handle email and scheduling for busy business owners.
Why it diedIt hired many assistants as employees without the cash to pay them and closed in the middle of its expansion.
The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
Answerly and Vanti tried to hold customer conversations with software that could not understand free speech or varied wording, and Zirtual used people, whose wages grew with every customer. Estera uses current language and speech models, which the co-founder told ZF are good enough in Romanian that most callers do not notice they talk to an AI.
The cost structure also changed: the model providers charge per use, so Estera can sell plans by minutes and let customers build the agent themselves. The competition problem remains, since many voice-agent companies sell the same capability.
Can one person copy Estera now?
Yes, for a language market the large vendors serve poorly.
Voice and WhatsApp agents are built from models and phone services that anyone can rent, so the product is easy to copy. The co-founder told ZF that anyone can build such a solution and that distribution is what matters; Estera's bet is one country and its language first.
Where it is still open: Receptionist agents for other Eastern European and smaller European languages, and for single industries such as veterinary clinics or driving schools.
- 1Pick a region or language where the funded voice-agent companies do not sell yet, and test the models' quality in that language.
- 2Let customers build and test an agent free and pay only when it goes live on their own number.
- 3Remove the setup call: a guided three-minute setup from the customer's website and documents.
- 4Start with an industry where results show in one month, such as clinics that count booked appointments.
What could go wrong for Estera?
The risks are the mirror image of the causes above.
What is not verified
No revenue figure was found. The only traction numbers are what co-founder Andrei-Fredy Crăciun told ZF on 6 August 2026 (more than 700 users, 20 paying, more than 6,400 calls and messages), and he said the team's internal target before fundraising was to pass EUR 10,000 in monthly recurring revenue, which suggests it was below that then. No TrustMRR profile or other payment-processor profile was found. The terms, privacy policy and data processing agreement are on app.tryestera.com, whose robots.txt disallows ClaudeBot, so they were not read. A reader could look for a payment-processor profile or a later interview with revenue.
Outbound WhatsApp messages to lists of property owners
Estera Messages, a limited-access tool for real estate agencies, sends "personalized WhatsApp outreach to property owners" from lists that the agency uploads or imports "from real estate portals". The company's FAQ says it "operates through the official WhatsApp API" and that it helps customers "build compliant opt-in flows and message templates". The pages read do not say how the people on such lists agreed to be contacted.
Callers may not know they talk to an AI
The Voice FAQ says: "Most callers don't realize they're speaking to an AI unless told". The co-founder told ZF that in about 95% of cases callers cannot tell. The pages read do not say whether the agent says it is an AI at the start of a call. The product uses stock synthetic voices ("choose from multiple voice styles"); no page offers cloning of a real person's voice.
Trust signals that do not match
The home page counter ("5,200+ conversations and calls handled", November 2026) is lower than the 6,400 the co-founder gave ZF three months earlier, and the "€210,000+ recovered" figure has no stated method. The pricing page shows a Trustpilot rating of 4.3 that could not be checked, since Trustpilot does not allow Claude's crawler. One of the four case studies is about an agency whose managing director shares a co-founder's surname; no source says whether they are related. The clinic case study says the agent can suggest an exam type from a patient's symptoms.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- C Estera home page, downloaded 5 November 2026Read in full, including the FAQ answers in the page source. Source for the product, the counters, the integrations, the legal entity and the press links.
- C Estera pricing page, downloaded 5 November 2026Read in full, including the plan table, the FAQ in the page source and the payment line.
- C Estera About page, downloaded 5 November 2026Read in full. Source for the two co-founders and their roles.
- C Ziarul Financiar, 6 August 2026, by Ioana Nita: ZF IT Generation. Start-up Pitch. Andrei-Fredy Crăciun, cofondator, Estera AI - recepţioneri AI pentru IMM-uri: Vrem să atragem peste 1 milion de euro la o evaluare de 5 milioane, după ce trecem de 10.000 de euro venituri lunare recurenteRead in full, in Romanian. A journalist's article built on the co-founder's statements; source for the users, paying customers, start date, models, funding plans and markets.
- C Comunicate Wall-Street.ro, June 2026: Estera, platforma care automatizează apelurile și conversațiile companiilor cu AI, fondată de doi tineri româniRead in full, in Romanian. A company press release on a release platform (page metadata dated 22 June 2026); source for the founders' earlier-products claim.
- C Estera Messages page, downloaded 5 November 2026Read in full, including the FAQ answer on WhatsApp policies in the page source.
- C Product Hunt product page for Estera, downloaded 5 November 2026, and Product Hunt's official API record for post 1205300Read the launch section (day rank, points) and the makers in the API record, which gives the launch date 27 July 2026.
- C Estera Voice page, downloaded 5 November 2026Read in full, including the FAQ answers in the page source on voices and AI disclosure.
- C Estera case studies: Remenix, Neocasa, SkyLine Properties and Eva Estate, downloaded 5 November 2026Read the four case-study pages in full. Vendor-written, quoting customers.
- A Federal Reserve statistical release H.10, Foreign Exchange Rates, released 2 November 2026Read the euro row: US$1.1163 per euro on 30 October 2026, used for the price conversion.
- C Database of failed startups, local exportLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
Run this company? Send corrections or a verified figure to aiolum@livyden.com