Metorial: Product, team and how it was built
Metorial (Open-source MCP and skills gateway for AI agents, Austria): 3.4K GitHub stars and paid plans live, November 2026. Evidence grade E: no revenue figure was checked. Team: 2 co-founders; Y Combinator lists a team of 4 in Vienna in November 2026. Launched 7 October 2025; still running.Added 5 November 2026 · updated 6 November 2026
- Outcome
- 3.4K GitHub stars and paid plans live, November 2026
- Evidence
- E no revenue figure was checked
- Launch
- 7 October 2025
- Status
- Still running
- Team
- 2 co-founders; Y Combinator lists a team of 4 in Vienna in November 2026
- Outside funding
- Y Combinator (Fall 2025); amount not published
- Main channel
- Not stated
What is not verified
No revenue figure for Metorial was found. We looked for a payment-processor profile, a founder interview with figures, a funding announcement with a round size and a filing. The traction claims are the company's own: "Used by engineers at FAANG, the Big 4, and American Express" and "piloting with leading startups and Fortune 500s" (launch post, 7 October 2025), and "working with startups, unicorns, and fortune 500 companies" (launch post, 12 November 2025). The home page shows five company names under "Trusted by companies like" without dates or case studies for most of them; these were not checked with the companies. GitHub stars show developer interest, not paying customers. A reader could look for an Austrian company register entry, a funding announcement or an 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.
Metorial is an MCP gateway. MCP, the Model Context Protocol, is a standard way for an AI model to call an outside tool, such as "search Notion" or "update a Salesforce deal"; an MCP server offers those tools, and a gateway is one service in front of many servers that handles sign-in, access rules and logging. Its home page says it offers "1,100+ MCP servers and 21,000+ tools" behind one link per person, and it checks each call for prompt injection, text hidden in data that tries to make the model do something its user did not ask for.
Tobias Herber (CTO) and Karim (Wen) Rahme (CEO) founded it in 2025; Y Combinator lists 4 people in Vienna, Austria, and the Fall 2025 batch. Its launch post of 7 October 2025 said the code was "Starred by more than 3000 GitHub devs"; the GitHub page showed 3,374 stars in November 2026. The pricing page lists a free plan and a Scale plan that it prices at US$250 a month.
No revenue figure was found, so this story has grade E. The page shows what the product does, its prices, and earlier integration platforms that failed.

How much revenue does Metorial 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 |
|---|---|---|---|---|
| Revenue of Metorial | not published | C | No payment-processor profile, founder interview with figures, funding announcement or filing with a revenue figure was found. See "What is not verified". | [1, 2, 3] |
| GitHub stars, November 2026 | 3,374 stars, 371 forks | C | GitHub page of metorial/metorial, read 5 November 2026. Stars are a sign of developer interest, not of paying customers. | [6] |
| GitHub stars, October 2025 | "more than 3000" | C | Company's launch post of 7 October 2025: "Starred by more than 3000 GitHub devs." | [2] |
| Founded, team and location | founded 2025; 4 people; Vienna, Austria | C | Y Combinator company page, downloaded 5 November 2026. It also lists 3 open engineering roles. | [1] |
| Founders | Karim (Wen) Rahme (Founder/CEO) and Tobias Herber (Founder/CTO) | C | Y Combinator company page. The second launch post begins "We're Wen and Tobias". | [1, 3] |
| Funding | Y Combinator, Fall 2025 batch; amount not published | C | Y Combinator company page. No round size was seen. | [1] |
| Catalog, November 2026 | "1,100+ MCP servers and 21,000+ tools" | C | Home page. The October 2025 launch post said 600+ MCP servers; the GitHub description says 1200+ integrations. | [4, 2, 6] |
| Prices, November 2026 | Dev free; Scale US$250 a month; Enterprise custom | C | Pricing page. Scale includes 20 Workforce seats; the page prices extra seats at US$20 each. | [5] |
| Customer claims | "Used by engineers at FAANG, the Big 4, and American Express" | C | Launch post of 7 October 2025. The home page shows the names Super.com, Tractive, Crunched, Moonfire and Cranston under "Trusted by companies like" and one named testimonial from Moonfire. None of these were checked with the companies. | [2, 4] |
| Security claims | SOC 2 Type II and GDPR, EU and US regions | C | Home page FAQ. The audit report itself was not seen. | [4] |
| Code license | GitHub shows "Other"; the catalog is Apache License 2.0 | C | GitHub page. The home page calls the product "open core", so some parts may be closed. | [6, 4] |
| Time from the first launch post to the star count | 13 months | calc | Calculated: October 2025 to November 2026. | [2, 6] |
How to read this table: a grade C value is the company's own statement, on its site, its GitHub page or its Y Combinator posts. A "calc" value was computed on this page. No value is grade A or B, and no revenue figure was found.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Dev | 0 | 500,000 tool calls a month, up to 10 integrations, 2 seats | |
| Scale | 250 | 2.5 million tool calls a month, 20 Workforce seats, unlimited integrations | |
| Extra Workforce seat (Scale) | 20 | per seat | |
| Enterprise | custom | custom | on-premises, SAML SSO, RBAC, SLA |
The page shows monthly prices only and no yearly discount. The Scale price is shown for 20 seats, which it includes; the page lists each extra Workforce seat at US$20. The free Dev plan allows 2 team members, 10 provider integrations and one project. Enterprise adds on-premises deployment, SAML single sign-on and role-based access control, at a price set by contract. The code can also be self-hosted from GitHub.
When did Metorial launch, and how did it grow?
From an open-source repository in April 2025 to a governed gateway for company agents, in about 19 months. Each entry carries its grade.
- 14 April 2025CThe metorial/metorial repository is created on GitHub. [6]
- 7 October 2025CLaunch post on Y Combinator, "Metorial: Vercel for MCP": 600+ MCP servers deployable in three clicks, OAuth handled, SDKs, and more than 3,000 GitHub stars. [2]
- 12 November 2025CSecond launch, Magic MCP: one link that connects Cursor, Claude Code or Claude Desktop to 600+ integrations with no configuration. [3]
- Fall 2025CY Combinator batch; the company page lists 4 people in Vienna. [1]
- November 2026CThe home page sells a governed gateway for company agents: per-person access, prompt-injection checks, audit logs, agents in Slack and Teams. Free and US$250 plans on the pricing page; 3,374 GitHub stars. [4, 5, 6]
Why did Metorial work?
Four causes, in the order they mattered. Each one is a reusable lesson.
Timing
The founders saw MCP spread as the standard for agent tools and built the missing production layer early: hosting, sign-in and logs. The launch posts sell speed, "hours, not weeks", to teams that wanted agents in production.
Open code as the entry point
The code is on GitHub, and the first launch post leads with the star count. Developers can self-host it for free; companies that want the hosted version, more tool calls or compliance features pay.
Sell control to the company, not the model
By 2026 the pitch is governance: which person may let which agent use which tool, with a record of every call. The product works with any MCP client, so it does not depend on one AI chat app winning.
How it was built
These details come from the company's GitHub README, its launch posts and its own pages. They are grade C. [6, 2, 3, 4]
- Most of the code is TypeScript on the Bun runtime; the MCP engine is written in Go, according to the README.
- PostgreSQL stores data, Redis handles caching and real-time data, and MongoDB stores usage data and logs. The dashboard is built with React.
- MCP servers run in Docker containers. The launch posts describe a serverless runtime with sub-second cold starts, per-user isolation and stateful hibernation for long connections.
- SDKs for TypeScript and Python, with examples for the Vercel AI SDK, OpenAI, Anthropic, Google Gemini and LangChain.
- Credentials are kept in a vault, calls run in an isolated enclave, and a component called Protoguard screens inputs for prompt injection, according to the home page.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Metorial, and why did they fail?
Three companies in our database of failed startups offered one layer that connects developers or teams to many outside services, and failed. The values below are from that database and are not independently verified.
Singly
The promiseOne API to reach user data across many social networks and online services.
Why it diedIt depended on other companies' APIs, which changed and limited access, while Zapier and IFTTT offered broader automation.
Snapboard
The promiseOne dashboard where developers and teams connect their APIs and tools.
Why it diedZapier and Microsoft Power Automate offered the same connections, and it could not stand out.
Piston Cloud
The promiseEasy private cloud infrastructure for enterprises, deployed in their own data centers.
Why it diedEnterprises moved their infrastructure to AWS, Google Cloud and Azure instead.
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
Singly and Snapboard sold one layer in front of many outside services and lost when larger platforms offered the same connections, or when the services changed their APIs. Piston Cloud sold enterprise infrastructure that the big clouds then offered as a service. Metorial faces both risks: AI clients and cloud providers can add their own tool catalogs and access rules.
Two things differ. MCP is a shared standard, so one integration works in Claude, ChatGPT, Cursor and other clients, and the integration work is not tied to one platform. And the buyer is a company that wants one place to decide which agent may act for which employee, a problem that a single AI client cannot solve across all the others.
Can one person copy Metorial now?
Hard for one person.
Enterprise buyers ask for SOC 2 audits, single sign-on, on-premises installs and support agreements, and the product runs a large catalog of servers. A solo founder could copy a small part: a hosted MCP server, or a gateway for one kind of tool or one industry.
Where it is still open: MCP servers and gateways for one regulated industry or one software stack, where the access rules matter more than the size of the catalog.
- 1Pick one set of tools that agents in one kind of company need, and host MCP servers for them with sign-in handled.
- 2Publish the code, so developers can try it before they pay.
- 3Log every tool call by person and agent, because that is what a company's security team asks for first.
- 4Put the free and paid limits on a public pricing page from the start.
What could go wrong for Metorial?
The risks are the mirror image of the causes above.
What is not verified
No revenue figure for Metorial was found. We looked for a payment-processor profile, a founder interview with figures, a funding announcement with a round size and a filing. The traction claims are the company's own: "Used by engineers at FAANG, the Big 4, and American Express" and "piloting with leading startups and Fortune 500s" (launch post, 7 October 2025), and "working with startups, unicorns, and fortune 500 companies" (launch post, 12 November 2025). The home page shows five company names under "Trusted by companies like" without dates or case studies for most of them; these were not checked with the companies. GitHub stars show developer interest, not paying customers. A reader could look for an Austrian company register entry, a funding announcement or an interview with figures.
Competition
The first launch post calls Metorial the "only" serverless MCP platform with sub-second cold starts, OAuth and per-user isolation; that is the company's own claim. On this site, LLM Gateway sells a gateway for AI model calls and Advanced GSC MCP sells one hosted MCP server, while Metorial sells the layer between agents and many company tools.
Dependence on AI clients and model providers
Metorial does not make a model. Its gateway is used through AI clients made by others; the home page names Claude, ChatGPT, Cursor and Codex, and the README gives examples for the OpenAI, Anthropic and Google Gemini APIs. The home page's sample cost view counts "model and tool usage" per session. If those clients add their own access rules and tool catalogs, part of the gateway's value moves to them; supporting any MCP client is the company's answer.
Open core and the license
The home page says the product is "open core", and GitHub shows the repository license as "Other" while the catalog is under the Apache License 2.0. Readers who plan to self-host should read the license file for which parts may be used commercially.
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 Y Combinator company page: MetorialRead in full; downloaded by the pipeline on 5 November 2026. Source for the founders, founding year, team size, location, batch and open roles.
- C Y Combinator Launches, 7 October 2025: Metorial: Vercel for MCPRead in full. Company's first launch post. Source for the 600+ servers, the star count, the customer claims and the founders' story.
- C Y Combinator Launches, 12 November 2025: Magic MCP by Metorial - 600+ integrations for vibe coding, zero config requiredRead in full. Source for Magic MCP, the serverless runtime and the customer statement.
- C Metorial home page, read 5 November 2026Read in full. Source for the gateway features, catalog size, logo wall, testimonial, FAQ, compliance claims and supported clients.
- C Metorial pricing page, read 5 November 2026Read in full. Source for the plans, limits and seat price.
- C GitHub: metorial/metorialRead the repository page and README, 5 November 2026. Source for the stars, forks, creation date, license and tech stack.
- 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