Aiolum Browse stories
All stories
AI moderation that hides toxic and hateful comments under social media posts and drafts replies for social media teams · Slovakia

TrollWall AI: Product, team and how it was built

Small teamFundedB2BSocial media

TrollWall AI (AI moderation that hides toxic and hateful comments under social media posts and drafts replies for social media teams, Slovakia): Clients in more than 10 countries, April 2026. Evidence grade E: no revenue figure was checked. Team: 4 co-founders (Tomáš Halász, Filip Strýčko, Pavol Bujňák, Tomáš Koctúr) who built it in 2022 and incorporated in March 2023; more people hired in 2024 after the first round; headcount today not published. Launched March 2023; still running.Added 6 November 2026

Outcome
Clients in more than 10 countries, April 2026
Evidence
E no revenue figure was checked
Launch
March 2023
Status
Still running
Team
4 co-founders (Tomáš Halász, Filip Strýčko, Pavol Bujňák, Tomáš Koctúr) who built it in 2022 and incorporated in March 2023; more people hired in 2024 after the first round; headcount today not published
Outside funding
EUR 500K from CB ESPRI Impact One in December 2023 and an EUR 800K seed round in April 2026 led by Seed Starter of Slovenská sporiteľňa with Slovak Investment Holding and angel investor Miroslav Fedor, as ITKeyMedia reported
Main channel
Not stated
No revenue figure

What is not verified

No revenue figure was found. The company publishes prices but no revenue, no payment-processor profile was found, and the press reports give funding, not sales. The Slovak company register sites were not read because finstat.sk disallows Claude agents on company pages. The client and profile counts are the company's own. A reader could look for the company's annual accounts in the Slovak register of financial statements, or an investor interview with figures. 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.

TrollWall AI does comment moderation: deciding which comments under a post stay visible. It uses a classifier, a machine-learning model that puts each comment into a category such as toxic, spam or acceptable, and then calls the social network's interface to hide the comment. The company's MCP page says comments are hidden "without notifying the author", and its FAQ says a client can approve a wrongly hidden comment again.

Tomáš Halász was a social media manager in 2022 and helped run an initiative for refugees from the war in Ukraine, where volunteers deleted hateful comments by hand. With three friends from IT projects, a CTO, a COO and a data scientist with a PhD and years of language-processing experience, he built a first version for internal use. They incorporated the company in March 2023 while still in their jobs, ITKeyMedia reported in May 2026. ITKeyMedia also reported a first investment of EUR 500K in December 2023 and an EUR 800K seed round in April 2026. Startitup wrote in April 2026 that clients in more than ten countries use it.

No revenue figure was found, so the evidence grade is E. What is checked is the founding team, the dates, the funding as reported in the press, the published prices and the client types the company and the press name.

TrollWall AI: Clients in more than 10 countries, April 2026. AI moderation that hides toxic and hateful comments under social media posts and drafts replies for social media teams, Slovakia, team of 2 to 10, evidence grade E.
The story in one card. Share it with the address of this page.
01

How much revenue does TrollWall AI 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.

An AI comment-moderation service from Slovakia started by four co-founders, funded twice and sold per social profile; no revenue figure is published
FactValueGradeHow we knowSources
Clients, April 2026clients in more than ten countriesCStartitup, 9 April 2026: "Už dnes totiž riešenie využívajú klienti vo viac ako desiatich krajinách" (translation: "Clients in more than ten countries already use the solution today"). The company's press page quotes the same words from the investor release.[5, 7]
Founding team4 co-founders: Tomáš Halász (CEO), Filip Strýčko (CTO), Pavol Bujňák (COO), Tomáš Koctúr (data scientist who led the detection layer)CITKeyMedia, 13 May 2026. World In You (July 2025 interview) says Halász began "with an IT expert friend".[4, 6]
Startfirst version built in 2022 for internal use; company incorporated in March 2023CITKeyMedia, 13 May 2026; the about page says "It all began in 2022".[2, 4]
FundingEUR 500K (December 2023) and EUR 800K seed (April 2026)CITKeyMedia reported both rounds; Startitup reported the EUR 800K round on 9 April 2026 with quotes from Seed Starter and Slovak Investment Holding.[4, 5]
Profiles and countries, July 2025hundreds of social media profiles across seven countriesCFounder biography on the World In You interview page, July 2025. The figure is the company's own and not dated more exactly.[6]
Free use by NGOs, July 2025more than 50 organisationsCWorld In You interview: the service is free for smaller NGOs, "currently supporting over 50 organizations".[6]
Languages14 languages (May 2026); the home page says "19+ Languages"CITKeyMedia counts 14 including three variants of Spanish. The home page headline says 19+ but its own list names 12 languages.[1, 4]
PlatformsFacebook, Instagram, TikTok, YouTube; developer API and MCP serverCHome page and MCP page, read 6 November 2026.[1, 3]
Pricesfrom EUR 39 to EUR 449 per profile a month, Enterprise from EUR 499 a monthCThe company's MCP pricing page, 6 November 2026; launch prices for accounts created in 2026.[3]
Months from incorporation to the April 2026 client count37 monthscalcCalculated: March 2023 to April 2026.[4, 5]

How to read this table: a grade C value is what the company states on its own pages or what the press reports from the company and its investors. A "calc" value (grade X) was computed on this page. No value comes from a payment processor, a filing or a journalist's own check.

TrollWall AI MCP pricing, read on trollwall.ai/mcp on 6 November 2026 (grade C); the main pricing page gives no prices and asks for a demo
PlanPer month, billed yearly (US$)Per month, billed monthly (US$)Limit
ConnectEUR 39 per profileEUR 39 per profile500 comments per profile; reports, like and hide through your agent; no automatic moderation
Standard Moderation 500EUR 99 per profileEUR 99 per profileup to 500 comments per profile; automatic hiding of toxic comments, drafted replies
Standard Moderation 1000EUR 249 per profileEUR 249 per profileup to 1,000 comments per profile
Standard Moderation 5000EUR 449 per profileEUR 449 per profileup to 5,000 comments per profile
Enterprisefrom EUR 499from EUR 499quoted by usage; autonomous moderation and reply agents, priority support

The MCP page says prices are per social profile, per month, based on comment volume, that every plan starts with one month free (up to 3 profiles, no credit card), and that these are launch prices guaranteed for 12 months for accounts created in 2026. Prices are in euros. The main pricing page lists a Starter plan (from 1,000 comments a month, 1 language) and a Pro plan (more languages, sentiment, reply suggestions, knowledge base) with no price, "including free trials and annual billing discounts". No yearly price is published, so both columns show the monthly price.

02

When did TrollWall AI launch, and how did it grow?

From hand-deleted comments in 2022 to a paid moderation service with agents and an MCP server in 2026. Each entry carries its grade.

  1. 2022CAfter the invasion of Ukraine, Tomáš Halász and volunteers delete hateful comments by hand; he and three friends build a first moderation model for Slovak for internal use. [2, 4]
  2. March 2023CThe four founders incorporate the company and keep their jobs; early users include heads of state, a telecom operator, brands and news media, ITKeyMedia reported. [4]
  3. December 2023CFirst investment, EUR 500K from CB ESPRI Impact One, as ITKeyMedia reported; the team works on it full time from 2024. [4]
  4. 2024CThe company lists the Startup Awards Slovakia win in the Society category and a Latitude59 pitch competition final. [2, 5]
  5. April 2026CEUR 800K seed round led by Seed Starter of Slovenská sporiteľňa with Slovak Investment Holding and angel investor Miroslav Fedor, as Startitup and ITKeyMedia reported; clients in more than ten countries. [4, 5]
  6. November 2026CThe site sells moderation and reply agents and an MCP server for AI agents, with published per-profile prices and a one-month free trial. [1, 3]
03

Why did TrollWall AI work?

Four causes, in the order they mattered. Each one is a reusable lesson.

01

One model per language and country

Halász told Startitup that the company does not train one universal model but adapts it to each country, with "viacero verzií rovnakého jazyka" (translation: "several versions of the same language"). ITKeyMedia reported that it works with external linguistic agencies and native speakers, which is slower but catches local slang and irony.

02

Hide, do not delete, and leave the decision to people

The product hides comments without notifying the author, so the author is not provoked to post again. Halász told Startitup that the final decision must stay with a person, and the team says it moderates hate and abuse, not opinions or criticism.

03

Sell to accounts that attract abuse

The client list named in the press, heads of state, political parties, a football club, a telecom operator, news media and an ice hockey championship, are accounts where abuse is constant and visible. The price is per social profile and comment volume, so a busy account pays more.

How it was built

These details come from the company's own pages and from interviews the founder gave. They are grade C. [1, 3, 4]

  • The core hate detection is "fully developed in-house by the founders", Halász told ITKeyMedia in May 2026, and he said it outperforms generic GPT-based models in moderation quality.
  • Models are trained with native speakers and external linguistic agencies for each language; Spanish has three variants for Latin American markets.
  • Connections to Facebook, Instagram, TikTok and YouTube, a developer API, and an MCP server that the MCP page shows with Claude and ChatGPT as clients.
  • Reply drafting, sentiment analysis and a customer knowledge base are part of the Pro plan; no downloaded page names the provider of the model that writes the replies.

Tools named in the sources

Each one opens its page in the Tools tab, with every story that uses it.

04

Who tried this before TrollWall AI, and why did they fail?

Three companies in our database of failed startups tried to manage comments or social media accounts for brands and failed. The values below are from that database and are not independently verified.

Died 2007 to 2016Burned 67 (database value; units not given)

Livefyre

The promise

Real-time comment and social content tools that turned website comment sections into filtered conversations.

Why it died

Disqus and Facebook Comments offered simpler integrations, and after Adobe bought it the product was absorbed into Adobe's suite.

Competition
Died 2011 to 2014Burned US$610K

PostRocket

The promise

Data-driven timing of Facebook page posts to increase their reach.

Why it died

Facebook changed its API and algorithm, and the company had no other platform to fall back on.

Product/Tech Failure
Died 2010 to 2017Burned US$1M

ThinkUp

The promise

Personal analytics for social media accounts beyond likes and retweets.

Why it died

The platforms built their own analytics and restricted API access, and larger suites such as Hootsuite and Buffer covered more needs.

Competition

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

Livefyre ran its own comment system and lost to the free comments of large platforms. TrollWall AI does not replace the platform's comments; it works on top of Facebook, Instagram, TikTok and YouTube and only decides which comments to hide.

PostRocket and ThinkUp show the risk of living on one platform's API. TrollWall AI spreads over four platforms and a developer API, and language models made it practical to add drafted replies and sentiment reports on top of the classifier.

05

Can one person copy TrollWall AI now?

Not alone, and not quickly.

TrollWall AI needed four founders with IT, infrastructure and language-processing skills, its own trained models for many languages, work with native-speaker linguists, and two outside rounds. One person could copy a narrow version: comment moderation for one language or one type of account, built on a general model.

Where it is still open: Moderation for one under-served language, or for one kind of account such as local sports clubs, schools or town councils, at a price below agency rates.

  1. 1Pick one language or region where the large platforms moderate poorly.
  2. 2Collect labelled examples of abuse with native speakers and test a classifier against them.
  3. 3Hide rather than delete, and show the client what was hidden so they can check it.
  4. 4Price per profile and comment volume, and start with accounts that receive abuse every day.

What could go wrong for TrollWall AI?

The risks are the mirror image of the causes above.

What is not verified

No revenue figure was found. The company publishes prices but no revenue, no payment-processor profile was found, and the press reports give funding, not sales. The Slovak company register sites were not read because finstat.sk disallows Claude agents on company pages. The client and profile counts are the company's own. A reader could look for the company's annual accounts in the Slovak register of financial statements, or an investor interview with figures.

Political clients and hidden comments

The home page says the product is "Trusted by Head of States, political parties", a For Politics page sells it to "political organizations, campaigns, and movements" with a testimonial from the social media manager of Petr Pavel's presidential campaign, and the careers page names political organizations as target clients; ITKeyMedia names a former Slovak president, the Czech president and political parties as users. The For Politics page says the agents "hide and moderate hate, misinformation, and manipulation" and calls this "Protect Your Narrative", and the FAQ says the product also hides "disinformation links, and any custom keywords you define in your blacklist". The MCP page says comments are hidden "without notifying the author". ITKeyMedia reported the team's view that "Legitimate criticism, political disagreement or sensitive discussion is not the target", and that the team sometimes pushes back on a client's definition of harmful content. These statements are the company's own; no downloaded page shows an outside check of what is hidden on political accounts.

Team and funding

ITKeyMedia names four co-founders and says the company hired more people in 2024 after the first round. No downloaded page gives today's headcount; the careers page says "Currently no open positions" and also shows a sales manager role for Poland. The company raised EUR 500K in December 2023 and EUR 800K in April 2026, as ITKeyMedia and Startitup reported.

Badges and counts

The home page shows a "Trusted by" logo wall and the about page lists awards; the logos were not checked one by one. The about page lists the Latitude59 2024 pitch competition as "Top 10 Finalist"; no outside page was checked for it. The home page counters say 52,000,000+ comments checked and 13,500,000+ toxic comments hidden, and the home page says clients in 17 countries while the For Politics page says 16; these counts are the company's own and were not checked. The home page says "19+ Languages" but lists 12, and ITKeyMedia counts 14.

Competition and model providers

On this site, Watchdog sells an AI moderation bot for Discord, Telegram, WhatsApp and Reddit communities; TrollWall AI works on brand and public-figure accounts on Facebook, Instagram, TikTok and YouTube. Halász told ITKeyMedia the hate detection is developed in-house, and World In You wrote in 2025 that the model is "built upon open-source NLP models, trained with proprietary data", so the core does not depend on a paid model provider; the product does depend on the four platforms' interfaces, and no page names the provider of the reply-drafting model.

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.

  1. C TrollWall AI home page, read 6 November 2026
    Read in full. Source for the product, platforms, languages and client types.
  2. C TrollWall AI about page, read 6 November 2026
    Read in full. Source for the 2022 start, the CEO's title and the awards list.
  3. C TrollWall AI pricing page and MCP page, read 6 November 2026
    Read both in full (the main pricing page at /pricing). Source for the per-profile prices, the free trial and hiding without notifying the author.
  4. C ITKeyMedia, 13 May 2026: TrollWall AI Scale Human-Centered Moderation Technology with EUR 800K Seed Funding
    Read in full. Source for the four co-founders, the March 2023 incorporation, both funding rounds, the in-house models and the languages.
  5. C Startitup, 9 April 2026: Slovenský startup získal 800-tisíc eur: Chce bojovať proti nenávisti na internete
    Read in full (Slovak). Source for the seed round, the investors, the clients in more than ten countries and the one-model-per-country approach.
  6. C World In You, interview with Tomáš Halász, July 2025
    Read in full. Source for the founding with an IT expert friend, the client types, the free NGO program and the profile count.
  7. C TrollWall AI press page, read 6 November 2026
    Read in full. Source for the list and dates of press coverage and the investor release text.
  8. C TrollWall AI careers page, read 6 November 2026
    Read in full. Source for the Polish sales role and its target clients.
  9. C Database of failed startups, local export
    Local export. AI-written summaries, not independently verified. Source for the failed predecessors.
  10. C TrollWall AI For Politics page and FAQ, read 8 November 2026
    Read in full (the FAQ at https://trollwall.ai/faq also read in full). Source for the political client types, the hiding of misinformation and disinformation links, and the custom keyword blacklists.

Run this company? Send corrections or a verified figure to aiolum@livyden.com

Back to all stories