NoimosAI is an autonomous marketing platform that runs a team of AI agents across social media, search, websites and, soon, customer email. On 15 September 2026 it added a Customer Engagement Agent that turns a written instruction into a triggered email sequence. The same afternoon it published two new developer packages that let other AI tools call its agents. The company describes the whole product as “the world’s first all-in-one autonomous AI marketing team”.
That launch line is not new. The blog post “Introducing NoimosAI” is dated 17 January 2026 and says “Today, we are launching NoimosAI.” A second, larger launch followed on 10 June, when the founder told Japanese followers the platform had been fully released. Since then a new agent or feature has arrived every one to five weeks, and each one has been announced with a sponsored article on TestingCatalog.
This article is based on the NoimosAI website, pricing page, terms of use, privacy policy, help centre, changelog, npm packages and OAuth metadata, plus four Japanese press releases, the founder’s posts on X and all seven TestingCatalog articles about the product, all read on 15 September 2026. It explains what launched, who runs the company, how the multi-channel agents work, what the credits cost, and ten places where the company’s own pages disagree. It ends with alternatives and a pilot checklist.
Table of contents
- What NoimosAI Launched on 15 September
- Who Is Behind NoimosAI
- How the Autonomous Marketing Platform Works
- The Multi-Channel Agents, Channel by Channel
- NoimosAI Timeline: From Closed Beta to Full Release
- NoimosAI Pricing, Credits and Plan Limits
- Checking the Company’s Own Numbers
- Sponsored Coverage and a Blog Built for AI Search
- Where NoimosAI’s Own Pages Disagree
- What the Terms of Use and Privacy Policy Say
- Alternatives to Consider
- Who Should Try NoimosAI, and How to Pilot It
- FAQ: NoimosAI
- References
What NoimosAI Launched on 15 September
Two things shipped on 15 September. Neither appears yet on the NoimosAI homepage, its agents page or its product changelog, whose latest entry is dated 3 September.
The Customer Engagement Agent
TestingCatalog’s article went live at 16:12 UTC and is labelled Sponsored. It describes an agent that “assembles and runs lifecycle email flows from a written instruction”, and it is based on “material the company supplied” rather than hands-on testing. In the demonstrated example, a short request to onboard new users returns a complete sequence on an editable canvas.
The sequence starts with a trigger when someone signs up, sends a welcome email straight away, then checks whether the person replied. People who reply are routed to a booking step. People who stay quiet receive follow-up emails two and five days after the welcome. The whole flow reports against a product goal, which in the demo is the customer creating a first project.
Drafting, review and WhatsApp
Each step on the canvas can be opened and edited, and the same sequence can be described in plain text instead of built visually. The agent drafts the emails with brand colours, typography and layout, adds merge fields for customer data and previews the result on mobile and desktop. A review stage sits between generation and sending, so, in the article’s words, “nothing goes out unapproved.”
The material also mentions messaging channels “including WhatsApp”, and customer records that hold plan status, join date, last activity and a short activity history. That would let a trial account that has gone quiet act as a trigger. WhatsApp does not appear on the NoimosAI integrations page, and the company “has not published a release date or plan availability” for the agent.
Two developer packages the same afternoon
The npm registry shows a burst of releases in the hour before that article appeared. At 15:32 UTC the existing command-line package, @noimosai/cli, reached version 0.0.10. At 15:45 UTC two new packages were published, @noimosai/platform-cli and @noimosai/platform-mcp, described as a “Developer Platform” for “research, SEO, social data and media generation”. Both moved to version 0.1.1 within 20 minutes and carry the MIT licence.
The hosted connection already exists. A request to mcp.noimosai.com/mcp without a token returns HTTP 401 and points to OAuth metadata that supports dynamic client registration, PKCE and a single mcp:tools scope. The documentation says Claude, ChatGPT and other clients can “chat with the NoimosAI agent, list workspaces and integrations, and publish social media posts directly from inside the client”. We covered a similar pattern earlier today in our look at Showly, a host for AI agent-built web pages.
| Time (UTC) | Release | Source | What is missing |
|---|---|---|---|
| 15:32 | Command-line package reaches 0.0.10 | npm registry | No readme on npm |
| 15:45 | Platform CLI and MCP packages published at 0.1.0 | npm registry | Linked GitHub repository returns 404 |
| 16:02 to 16:04 | Both platform packages move to 0.1.1 | npm registry | No changelog entry |
| 16:12 | Customer Engagement Agent article | TestingCatalog, Sponsored | No release date or plan availability |
Who Is Behind NoimosAI
The product comes from a small company with a founder who is well known on Japanese social media. Its legal home is harder to pin down.
AGOS LABS and Kosuke Yokoyama
NoimosAI is built by AGOS LABS TECHNOLOGIES LTD. Its chief executive is Kosuke Yokoyama, who is also named in the privacy policy as both the company representative and its personal data protection manager. The Japanese press releases give agoslabs.xyz as the corporate website.
Yokoyama’s Japanese X account has more than 86,000 followers. Its biography summarises a media business sale, then a token listing at a ¥15 billion market capitalisation, then NoimosAI. The English-language account reads “college dropout, 2 exits”. TestingCatalog adds that the team “previously worked on Web3 products before transitioning to marketing automation.”
Dubai, Delaware and a Tokyo court
The company’s location depends on which page you read. In June 2025, a hiring post from the founder on X described the business as a UAE company developing NoimosAI as an SEO and GEO agent, alongside a social media agent called FotisAI. TestingCatalog later called AGOS LABS “a Dubai-registered company”. The help centre’s images are served from a documentation account named “agoslabstechnologiesfzco”, and FZCO is the suffix used by UAE free zone companies.
The terms of use say something different. They name AGOS LABS TECHNOLOGIES LTD. as “a Delaware corporation”, apply Delaware law and give “the Tokyo District Court” exclusive jurisdiction over disputes. The Japanese press releases list an address at 16192 Coastal Highway, Lewes, Delaware. A group with more than one entity could make all of this true at once, but no NoimosAI page explains the structure.
Not noimos AG of Zürich
A similar name belongs to an unrelated business. noimos AG, at noimos.ai, is a Zürich insurance technology firm that uses AI for motor claims, windscreen repair assessment and fraud prevention. Its site says it was founded in 2021 and has a team of about 30. Check the domain before you sign up: the marketing platform lives at noimosai.com.
| Field | Detail | Source |
|---|---|---|
| Operator named in the contract | AGOS LABS TECHNOLOGIES LTD., “a Delaware corporation” | Terms of use |
| Address in press releases | 16192 Coastal Highway, Lewes, Delaware 19958 | PR TIMES |
| Described elsewhere as | Dubai-registered, a UAE company, an FZCO | TestingCatalog, X, help centre |
| Chief executive | Kosuke Yokoyama | Press releases, privacy policy |
| Governing law and court | Delaware law, Tokyo District Court | Terms, articles 26 and 27 |
| Brand account on X | Created 1 November 2024, 2,660 followers | X |
| Domain registered | 16 May 2025 | Domain registry record |
| Different company, similar name | noimos AG, Zürich, motor insurance AI | noimos.ai |
How the Autonomous Marketing Platform Works
The product is easier to judge once you separate the idea, the parts and the setting that makes it autonomous.
Command marketing: goals in, finished work out
The January launch post calls the idea “Command Marketing”. Instead of operating software, a person gives a goal, and the agents build the strategy, assemble a team and do the work. The person becomes “the commander who makes the final strategic decisions”. Setup starts by connecting tools such as WordPress, social accounts and Google Analytics, which NoimosAI uses to build a knowledge base about the business.
Ten agents, a Feed and a Chat
At launch the company listed ten specialised agents: Growth Metrics and Strategy, Competitor Strategy, Social Listening, Industry News, Social Media, SEO, GEO, Event Outreach, Media Outreach and CVR Optimization. The agents page on the NoimosAI website still shows exactly those ten as “Available Now”. Work is requested in a Chat, finished output lands in a Feed for review, and Pages let users build dashboards, content calendars and inboxes from blocks, in a style the FAQ compares to Notion.
The help centre now frames this differently. Its capabilities page says the product “focuses on capability-based agent design rather than fixed agent categories” and lists eleven areas, adding Creative Production and Advertising. If you are comparing it with other AI employees and autonomous AI agents, the capability list is the better guide to what it does today.
Triggers make the agents autonomous
Autonomy comes from triggers. The documentation lists five trigger types: a schedule, a new Gmail message, a new post on a chosen X account, a new RSS item and a tracking link condition. Every paid plan allows 30 triggers. The sign-up trigger in the Customer Engagement demo is not one of the five documented types, which is another sign that the new agent is not yet generally available.
Memory you can no longer edit directly
Agents also draw on a knowledge base and a memory layer. The changelog entry for 17 August records “the removal of direct Memory editing”. Memory is now adjusted by what the documentation calls “24/7 memory agents”, using feedback in Chat and execution results. Users review or change memory by asking in Chat and approving a proposed change, rather than on a dedicated screen.
Advisor chooses the work
Advisor, released on 17 July, moves NoimosAI from doing assigned work to proposing it. It reads connected data such as recent social posts, Google Search Console and Google Analytics, compares performance with the account’s goals and queues recommended actions for one-click approval. The Japanese press release compares it with Claude Code, saying that just as coding agents turned engineers into reviewers, Advisor turns marketers into people who approve proposals.
The Multi-Channel Agents, Channel by Channel
“Multi-channel” is the core of the pitch, so it is worth checking which channels each part of the product covers. The answer changes from page to page.
Social Agent and its nine networks
The Social Agent went live on 7 August. TestingCatalog listed nine supported networks: X, Facebook, Instagram, Threads, TikTok, YouTube, Pinterest, Bluesky and Mastodon. It researches trends in a niche, studies high-performing posts, drafts carousels, product videos, user-generated-content-style videos and clips, then schedules them and reports on performance. It is aimed at solo operators and small teams who “often rely on intuition for content decisions”.
Creative Agent and generated media
The Creative Agent, announced on 30 June, studies competitor output and top-performing creatives on Meta, TikTok and LinkedIn, then produces assets built on the same patterns. The Japanese release lists full video generation, AI avatars that read scripts, image-to-video, voice cloning, generated music and clipping from long videos. The terms of use name one supplier: media features are “powered by third-party generation providers, including BytePlus”, the enterprise technology arm of TikTok owner ByteDance.
This kind of creative work depends on seeing what competitors run. For a sense of how that data is gathered, see our review of SpyTrend’s 1.1 billion-ad corpus inside AI assistants.
SEO Agent 2.0 and the Website Builder
SEO Agent 2.0 arrived on 12 August. It combines Semrush data on domain strength, keyword difficulty and backlink gaps with Google Search Console, GA4 and PageSpeed data. It then proposes work, such as closing a keyword gap a competitor already ranks for, and after approval can publish articles with images, internal links and meta tags to WordPress or to sites built inside NoimosAI.
The Website Builder, released on 10 July, builds a site from a chat, generates an llms.txt file and JSON-LD structured data, imports existing code from GitHub and sells domains inside the platform. The company calls it the “Self-Improving Web” because it reads Search Console and Analytics data and “rewrites its own code and HTML” to fix performance problems after launch.
Email, messaging and approvals from chat apps
Email currently runs through Gmail, which agents use to draft outreach and which can trigger an agent when a message arrives. LINE, Telegram and Discord appear on the integrations page as places to receive updates, approve tasks and assign work. The Customer Engagement Agent would add lifecycle email and WhatsApp messaging on top. Teams already automating inbox work may recognise the approach from Lindy’s CC-based meeting booking.
Listening, outreach and conversion
The remaining agents cover the rest of a small marketing department. Social Listening monitors brand mentions and potential leads, Competitor Strategy tracks rival traffic and keywords, Event and Media Outreach find speaking slots and journalists and draft pitches, and CVR Optimization analyses visitor behaviour on a site. Apollo appears as a prospect data source, which the Japanese launch release describes as holding lists of millions of companies and contacts.
| Channel | Social Agent article, 7 August | Integrations page | CLI publishing docs |
|---|---|---|---|
| X, Facebook, Instagram, Threads | Yes | Yes | Yes |
| TikTok, YouTube, Pinterest, Bluesky | Yes | Yes | Yes |
| Mastodon | Yes | Yes | No |
| LinkedIn (personal) | No | Yes | Yes |
| WordPress and note | Not listed | Yes | Yes, as articles |
| LINE, Telegram, Discord | Not listed | Yes, for approvals | No |
| Not listed | No | No |
NoimosAI Timeline: From Closed Beta to Full Release
The public record shows a product that has launched in stages over almost two years.
Account, domain and closed beta
The brand’s X account was created on 1 November 2024, and the noimosai.com domain was registered on 16 May 2025. In June 2025 the founder advertised for a TypeScript, React and Next.js engineer to join “a small global team” that was “still in an early phase”. A waitlist opened in early November 2025, and by mid-November the founder was inviting people into a free closed beta.
“Today, we are launching”
The launch post, “Introducing NoimosAI, The World’s First Autonomous AI Marketing Agent”, is dated 17 January 2026. In April the founder described that moment as a soft launch “three months” earlier and promoted a seven-day free trial. The terms of use and privacy policy were enacted on 26 April 2026, and the terms were amended on 5 June.
The June full release
On 10 June the brand’s X account posted “Introducing NoimosAI: The world’s first all-in-one autonomous AI marketing team”, a video post that had passed 333,000 views by 15 September. The founder’s Japanese post the same day said the product had been fully released. A PR TIMES release on 11 June announced full service for Japan, and TestingCatalog’s first sponsored article also appeared on 10 June.
A feature every few weeks
Since June the pace has been fast. The Creative Agent came on 30 June, the Website Builder on 10 July, Advisor on 17 July, the Social Agent on 7 August and SEO Agent 2.0 on 12 August. The first three were followed by Japanese press releases one to four days later. The Customer Engagement Agent and the developer packages followed on 15 September.
| Date | Event | Source |
|---|---|---|
| 1 Nov 2024 | Brand X account created | X |
| 16 May 2025 | Domain registered | Registry record |
| 7 Jun 2025 | Hiring post describes a UAE company | Founder on X |
| Nov 2025 | Waitlist, then free closed beta | Founder on X |
| 17 Jan 2026 | “Today, we are launching” | Company blog |
| 26 Apr 2026 | Terms and privacy policy enacted | Legal pages |
| 10 Jun 2026 | Full release and first sponsored article | X, TestingCatalog |
| 11 Jun 2026 | Japan launch press release | PR TIMES |
| 30 Jun 2026 | Creative Agent | TestingCatalog |
| 10 Jul 2026 | Website Builder | TestingCatalog, changelog |
| 17 Jul 2026 | Advisor | TestingCatalog |
| 7 Aug 2026 | Social Agent | TestingCatalog |
| 12 Aug 2026 | SEO Agent 2.0 | TestingCatalog |
| 17 Aug 2026 | Direct memory editing removed | Changelog |
| 18 Aug 2026 | Command-line package first published | npm |
| 15 Sep 2026 | Customer Engagement Agent and platform packages | TestingCatalog, npm |
NoimosAI Pricing, Credits and Plan Limits
Pricing is public and simple on the surface. The detail sits in the credits.
Three plans priced per user
There are three tiers, each priced per user per month. Pro costs $99 and targets solopreneurs, creators and freelancers. Team costs $249 and is aimed at small businesses and agencies. Advanced costs $499 for “high-volume marketers and demanding mid-market companies”. Every plan includes every agent, and each tier roughly doubles or triples the limits of the one below it.
| Limit | Pro, $99 | Team, $249 | Advanced, $499 |
|---|---|---|---|
| Monthly credits | 30,000 | 80,000 | 160,000 |
| Daily credits | 100 | 300 | 600 |
| Workspaces | 2 | 5 | 10 |
| Connected apps | 5 | 15 | 30 |
| Pages | 10 | 25 | 50 |
| Competitors and listening keywords | 3 each | 8 each | 16 each |
| Triggers | 30 | 30 | 30 |
| Knowledge base and media storage | 5GB each | 15GB each | 30GB each |
| Extra credits | $3 per 1,000 | $3 per 1,000 | $3 per 1,000 |
| Claimed equivalent team | $3,000 a month | $10,000 a month | $25,000 a month |
What a credit buys
The pricing FAQ defines a credit as “a unit of measurement for NoimosAI usage” and says different actions consume “different amounts of credits based on their computational complexity”. No rate card is published. Instead, each agent shows an “Estimated usage” figure, and the help centre warns that actual usage “can vary” with connected tools, data volume, trigger behaviour and prompt complexity. That makes a monthly bill hard to predict before you run real workflows.
Top-ups cost less per credit than Pro
A simple division shows an oddity. Pro’s 30,000 credits for $99 work out at 303 credits per dollar, while Team and Advanced both give about 321. Extra credits cost $3 per 1,000, which is 333 credits per dollar. Plans include features that top-ups do not, but on credits alone, buying more is slightly cheaper than the credits bundled into Pro.
The “marketing team” comparison
Each plan card claims to “unlock the output” of a marketing team costing $3,000, $10,000 or $25,000 a month. Divided by the plan price, that claimed value rises from about 30 times the price on Pro to 40 times on Team and 50 times on Advanced. The page gives no method for any of these figures, so treat them as marketing rather than a benchmark.
Free tier, free trial or free account
The pricing page has no free plan, only “Start for Free” buttons. The company’s llms.txt file invites visitors to “Create a free NoimosAI account”, and TestingCatalog’s 7 August article mentioned “a free tier available to start”. The founder, the Japanese press releases and a competitor’s comparison page all describe a seven-day free trial instead, and Capterra lists a free trial but no free version. The terms settle what matters most: unless you cancel before a trial ends, it “shall automatically renew and convert to a paid subscription”.
Checking the Company's Own Numbers
Most of the performance figures in circulation come from the company itself. Some simple checks help put them in context.
$1 million in annual recurring revenue in 30 days
TestingCatalog’s sponsored Social Agent article reported that Yokoyama “has publicly stated that the platform achieved $1 million in annual recurring revenue within 30 days of launch through organic distribution alone”. We could not locate the original post, and the claim does not say which of the launches it counts from.
Some arithmetic shows what it implies. $1 million a year is $83,333 a month. At list prices, that means at least 842 paid Pro seats, 335 Team seats or 168 Advanced seats, or a mix of the three, within a month. Discounts, annual billing and credit top-ups would change the mix, but not the scale.
A founder case study among the testimonials
The homepage section headed “Don’t Take Our Word for It” opens with a case study titled “Dominating EN & JP with Just 2 People: Proving 10x ROI via AI”. It reports more than 10,000 monthly visits, a tenfold return on investment and total reach above 20 million in one month. It is attributed to Kosuke Yokoyama, Founder and CEO, so it describes the company marketing itself. The Japanese launch release lists the same three figures among results achieved by users.
The other three testimonials give a first name or a single full name and a job title, but no company. None links to a business that a buyer could contact.
Beta results without customers
The January post listed three private beta results: marketing costs cut by “over $10,000” a month within weeks, organic traffic up “over 1000%” in two months using TikTok and X, and a Google ranking lifted from the second page to the top four. No customer, industry or starting baseline is named for any of them.
No independent reviews yet
Independent reviews are thin. The NoimosAI listing on Capterra showed zero user reviews on 15 September. The most detailed comparison we found is published by Runlo, a competitor, which is open about its interest and recommends itself for business-to-business software founders.
Sponsored Coverage and a Blog Built for AI Search
Much of what appears when you search for the product was paid for or written by the company. That is legal and common, but readers deserve to know it.
Seven articles, all labelled Sponsored
TestingCatalog has published seven articles about NoimosAI since 10 June, all under the same byline. Every one is labelled “Sponsored” at the top and bottom and carries a “Start testing” button in the middle. On the site the label is clear. Elsewhere it often disappears: Bing News lists the headlines alongside ordinary news, and an agent directory called MyAgentStack describes the product as “frequently cited in 2026 AEO/social buyer guides”.
| Date | Headline topic | Days since previous |
|---|---|---|
| 10 Jun 2026 | Launches autonomous AI marketing team | First |
| 30 Jun 2026 | Creative Agent for brand assets | 20 |
| 10 Jul 2026 | Website builder that automates SEO | 10 |
| 17 Jul 2026 | Advisor for automated marketing tasks | 7 |
| 7 Aug 2026 | Social Agent to automate content creation | 21 |
| 12 Aug 2026 | SEO Agent 2.0 for automated optimisation | 5 |
| 15 Sep 2026 | New Customer Engagement Agent | 34 |
Where the sponsored copy disagrees with itself
The articles are not consistent with each other. The 17 July piece says the product opened “to the public on June 10”. The 7 August and 15 September pieces say it opened in January. The 12 August piece says the Website Builder arrived first and “a Creative Agent covering social and ad assets followed shortly after”, yet the Creative Agent article is dated ten days before the Website Builder article.
144 blog posts, 119 of them rankings
The NoimosAI llms.txt file, a plain-text guide written for AI crawlers, lists 144 English blog posts. 119 of their titles contain “Best”, “Top” or “Alternatives”, such as “8 Best Sintra AI Alternatives in 2026”. Several open by naming NoimosAI as the pick, for example “The best Sintra AI alternative is NoimosAI for connected marketing strategy and execution”.
The robots.txt file explicitly welcomes eleven AI crawlers, including GPTBot, ClaudeBot and PerplexityBot, and llms.txt files exist in 11 languages. This is textbook generative engine optimisation, the discipline the GEO agent sells. It also means an AI assistant asked for the best autonomous marketing agent may be summarising the vendor’s own rankings.
How to read coverage like this
Sponsored articles and vendor rankings are normal marketing, and neither makes a product bad. They do change the question you should ask. Before relying on a claim, find out whether it comes from a customer, an independent tester or the company, and ask for a named business that will take a reference call.
Where NoimosAI's Own Pages Disagree
Early products change quickly, and some inconsistency is normal. These ten are worth knowing because they touch dates, legal identity, money, channels and approvals.
1. When the product launched
The blog post dated 17 January 2026 says “Today, we are launching NoimosAI.” TestingCatalog’s 17 July article says the product opened to the public on 10 June, the date the founder called the full release. The 7 August article went back to January.
2. Where the company is registered
The terms describe “a Delaware corporation” governed by Delaware law, with disputes sent to a Tokyo court. TestingCatalog calls it “Dubai-registered”. The founder’s 2025 hiring post describes a UAE company, and the documentation account is named after an FZCO.
3. How many credits each plan includes
The live pricing page lists 30,000, 80,000 and 160,000 monthly credits. The Capterra listing shows 15,000, 40,000 and 80,000 at the same prices, exactly half. It may simply be out of date, but a review site is often the first page a buyer reads.
4. Whether there is a free tier
One sponsored article promised “a free tier available to start”. The press releases and the founder describe a seven-day free trial. Capterra says there is no free version, and the terms describe a trial that turns into a paid subscription.
5. How many apps connect
The llms.txt file says “20+ platforms” and lists 20. The pricing page lists 21, including Google Docs. The integrations page lists 26, including Apollo, LINE, Telegram and Discord but not Google Docs. The help centre’s integration guide lists 20. The 15 September article adds WhatsApp, which appears in none of them.
6. Which social networks it can publish to
The Social Agent article lists nine networks, including Mastodon but not LinkedIn. The CLI publishing guide also lists nine, including LinkedIn but not Mastodon. The integrations page lists all ten.
7. Which languages it supports
The FAQ says the product is available in 12 languages, including Turkish and Slovak. The language-specific llms.txt files cover 11, including Swedish, but not Turkish or Slovak.
8. Which devices it runs on
The FAQ says NoimosAI “is currently designed for desktop use only”. The Capterra listing names Web, Android and iPhone or iPad as deployment options.
9. Whether anything goes out without approval
The homepage says users “approve executions with a click”, and the Customer Engagement material says “nothing goes out unapproved”. The MCP documentation’s post tool has a dryRun setting that “Defaults to false“, and warns: “Without dryRun, posts publish to the real social account.” The CLI also has a --now flag, and chat settings include an “approval mode”, so approval is a setting rather than a guarantee.
10. Which developer packages to install
The MCP and CLI guides tell users to install @agos-labs/noimosai-mcp and @agos-labs/noimosai-cli. On 15 September the npm registry returned “Not found” for both names. The packages that do exist are published under the @noimosai name, and they require Node.js 20 or 22 rather than the documented version 18.
| Topic | One source says | Another source says |
|---|---|---|
| Launch | 17 January 2026 | Public from 10 June 2026 |
| Registration | Delaware corporation | Dubai-registered, UAE company |
| Pro credits | 30,000 a month | 15,000 a month |
| Free access | Free tier | Seven-day trial, no free version |
| Integrations | 20 or 21 | 26 |
| Social networks | Nine with Mastodon | Nine with LinkedIn |
| Languages | 12, with Turkish and Slovak | 11, with Swedish |
| Devices | Desktop only | Web, Android, iPhone and iPad |
| Approval | Nothing goes out unapproved | MCP posts publish unless dryRun is set |
| Packages | @agos-labs names, Node.js 18 | @noimosai names, Node.js 20 or 22 |
What the Terms of Use and Privacy Policy Say
Few buyers read the contract before a trial. For an autonomous product that publishes on your behalf, it is worth ten minutes.
Delaware law, a Tokyo court and English that prevails
Article 26 applies Delaware law. Article 27 gives “the Tokyo District Court” exclusive jurisdiction, and users waive any objection to that forum. Article 29 says the English text prevails over translations. For a customer in the UK or the US, a dispute would therefore be heard in Japan under American state law. The page header says the terms took effect on 26 April 2026, while the closing provisions record an amendment effective on 5 June.
What the service is, according to the contract
Article 4 describes the service narrowly. It “enables Users to manage their SNS accounts collectively and suggests content for posting that is expected to achieve a high level of engagement”. It says nothing about SEO publishing, websites, outreach, email flows or autonomous execution, which are the features the marketing describes. Contracts often lag products, but buyers should know which description the vendor is bound by.
Trials, cancellation and publicity
Free trials convert to paid subscriptions automatically unless cancelled first. The company can end the agreement with 30 days’ notice, and a user who withdraws does so “on the last day of the current month” and forfeits any remaining term. Article 10 gives the company consent to use each customer’s name and logo in case studies and promotional material, and the terms describe no opt-out.
A clause that bans automation tools
Article 11 prohibits “Use of macros, automated functions, or tools that automate operations”, as well as scraping the service. The company presumably does not mean its own CLI and MCP server, but the clause does not carve them out. Teams planning scripts, scheduled jobs or third-party agents on top of NoimosAI should get written confirmation of what is allowed.
Liability capped at $100 or a year of fees
Article 14 supplies the service and all generated material “AS IS”, and makes the user “solely responsible for reviewing and verifying the accuracy and legality of all Generated Materials”. Article 15 caps the company’s liability at the greater of $100 or the fees paid in the previous 12 months, excludes indirect damages, waives class actions and requires users to indemnify the company. Article 28, the copyright notice clause, still contains the unfilled placeholder “[Contact Email]”.
Your data and AI training
The terms say the company “will not use Uploaded Data or Generated Materials to train, improve, or fine-tune its foundation AI models”, and the homepage says third-party AI providers never store or learn from customer information. The privacy policy says YouTube data may be sent to third-party language models to generate insights, and that the service runs on Google Cloud. It also charges ¥1,000 for some personal data disclosure requests.
The help centre describes a per-workspace setting called “Improve AI suggestions for all users”, without saying whether it is on by default. The terms add that generation activity is logged and may be shared with generation providers during compliance investigations. A cybersecurity or privacy review should check all three points before customer data is connected.
| Clause | What it says | Why it matters |
|---|---|---|
| Article 4 | Manages social accounts and suggests posts | Narrower than the product sold |
| Article 5 | Trials convert to paid automatically | Set a cancellation reminder |
| Article 10 | Company may use your name and logo | Ask for an opt-out in writing |
| Article 11 | No macros or automation tools | Confirm scripts and agents are allowed |
| Articles 14 and 15 | As-is service, liability capped | You carry the risk of bad output |
| Articles 26 and 27 | Delaware law, Tokyo District Court | Disputes are costly to bring |
| Article 28 | “[Contact Email]” placeholder | A sign of an unfinished contract |
Alternatives to Consider
An all-in-one agent team is one way to get marketing done with fewer people. It is not the only one.
Keep your scheduler and add AI slowly
Many small teams already use a social scheduler and a separate SEO tool. MyAgentStack frames the difference simply: “Buffer and Hootsuite schedule what you write. NoimosAI writes too.” If writing is not your bottleneck, a scheduler plus a few well-tested AI prompts may be cheaper and easier to control than handing over execution.
Runlo for business-to-business software
Runlo sells a narrower crew of agents for business-to-business software founders, built around Reddit, Search Console-driven SEO and LinkedIn prospecting. Its comparison page lists flat prices of $65 and $99 a month with no credit meter, and says nothing is published without the founder’s approval. It is a competitor’s own page, so verify its claims as carefully as any other vendor’s.
General-purpose agent builders
The NoimosAI blog’s own alternatives posts name Lindy, Relevance AI, Copy.ai and others, and describe Lindy as “a general-purpose AI teammate”. Tools in that group suit teams that want workflows beyond marketing, such as inbox handling, sales research or support, and are willing to configure more of the process themselves.
A human team or agency
For regulated products, brand-sensitive launches or businesses without time to review AI output every day, people remain the safer default. Our marketing services combine strategy with hands-on execution. Whoever you choose, ask for named clients and measurable results, which is the evidence missing from the public pages reviewed here.
Who Should Try NoimosAI, and How to Pilot It
The product is ambitious and moving quickly. Whether it is right for you depends on how much risk you can carry while it matures.
Good fits
Solo founders, creators and small teams who post across several networks and already rely on AI drafts are both the target and the best fit. Businesses marketing in Japanese and English may gain most, since the company markets heavily in both. A seven-day trial is long enough to connect two or three channels and judge the quality of the drafts.
Reasons to wait
Larger organisations have fair reasons to pause. The company’s legal identity is described three ways, disputes go to a Tokyo court, liability is capped at $100 or a year of fees, and the contract describes a much narrower service than the website. The newest agent has no release date, and there are no independent reviews or named customers to check.
Pilot checklist: access and money
- Start with one workspace and one or two channels.
- Set a reminder to cancel before the trial converts.
- Keep Advisor recommendations and chat in approval mode.
- Compare estimated and actual credit use for a full week.
- Keep triggers to a handful until output quality is proven.
Pilot checklist: control and evidence
- Test whether any path, including MCP and the CLI, can publish without approval.
- Ask in writing which legal entity you are contracting with.
- Check the “Improve AI suggestions for all users” setting.
- Request a written opt-out from publicity use of your name and logo.
- Ask for two named customer references in your sector.
FAQ: NoimosAI
What is NoimosAI?
NoimosAI is an autonomous AI marketing platform from AGOS LABS. A team of agents covers strategy, social media, SEO, GEO, outreach, websites and creative production, and connects to apps such as X, Instagram, TikTok, WordPress, Google Analytics and Slack.
What did NoimosAI launch on 15 September 2026?
It announced a Customer Engagement Agent that builds triggered email flows from a written instruction, described in a sponsored TestingCatalog article, and published two developer platform packages on npm. The agent has no published release date or plan availability.
How much does NoimosAI cost?
Pro costs $99 per user per month with 30,000 credits, Team costs $249 with 80,000 and Advanced costs $499 with 160,000. Extra credits cost $3 per 1,000. A seven-day free trial converts to a paid plan unless it is cancelled.
Does NoimosAI post without approval?
The web app is built around reviewing output before execution, and chat has an approval mode. However, the documented MCP post tool publishes immediately unless dryRun is set, so test every connection path before trusting it with a live account.
Does it work with Claude or ChatGPT?
Yes. A hosted MCP server uses browser sign-in, and the documentation gives setup steps for Claude Code, Claude.ai and ChatGPT. Its tools can chat with the agent, list workspaces and integrations, and publish posts. The documented local package for Cursor and VS Code could not be found on npm.
Is NoimosAI the same company as noimos AG?
No. noimos AG is a Zürich insurance technology company at noimos.ai. The marketing platform is at noimosai.com and is operated by AGOS LABS TECHNOLOGIES LTD.