The archive · Developer & Business Tools · Product decision · 2025–2026
Modelence bets one TypeScript stack can make vibe-coded apps ship; $3M YC seed
YC S25's Modelence sells AI builders a pre-wired TypeScript+MongoDB stack — auth, hosting, observability — betting vibe coding fails at the seams.
Modelence
What the business is
Modelence makes an open-source, all-in-one backend framework plus AI app builder: TypeScript and MongoDB with authentication, databases, hosting, cron jobs, monitoring and LLM tooling pre-integrated.
Starting capital：$3M seed led by Y Combinator, with Rebel Fund, Acacia Venture Capital Partners, Formosa VC and Vocal Ventures, announced 2026-01-28.
How it started
Co-founders Aram Shatakhtsyan (CEO) and Eduard built Modelence after watching AI write code but fail at deployment. Their diagnosis: the problem is not any single service but the stitching between Vercel, Supabase and the rest of the stack. Accepted into Y Combinator's summer 2025 batch, they launched an open-source TypeScript+MongoDB backend platform — framed as 'Supabase for MongoDB' — on Hacker News in August 2025.
What happened
Modelence announced a $3M seed on 2026-01-28, led by Y Combinator with participation from Rebel Fund, Acacia Venture Capital Partners, Formosa VC and Vocal Ventures. The company shipped an AI-native app builder on top of the framework that handles the full application lifecycle — APIs, databases and deployment — and its late-January 2026 Hacker News launch drew discussion of TypeScript guardrails that catch AI-generated type errors and agent-friendly auth and cron primitives.
How it ended up
Live and seed-funded: the framework is open source, the app builder is launched, and the company is positioning itself as an 'agent-native cloud' for the vibe-coding era as of early 2026.
Background
Modelence is a Y Combinator Summer 2025 startup founded by Aram Shatakhtsyan and Eduard on a simple observation: generative AI got very good at writing code, but AI-generated apps kept dying in the gaps between services. Their framework bundles authentication, databases, hosting, cron jobs, monitoring and LLM tooling into one TypeScript+MongoDB stack so a 'vibe-coded' prototype can reach production without engineers stitching Vercel to Supabase to everything else.
The company went public early through developers: an August 2025 Show HN post, titled 'Modelence – Supabase for MongoDB', asked the Hacker News audience to try the open-source framework, and the founders kept the repository public on GitHub. Inside YC's summer 2025 batch, the framework added an AI app builder that handles the entire application lifecycle — APIs, databases and deployment.
On 2026-01-28 Modelence announced a $3M seed round led by Y Combinator with Rebel Fund, Acacia Venture Capital Partners, Formosa VC and Vocal Ventures. TechCrunch framed it as an answer to the 'vibe-coding stack' problem: LLMs sped up coding, but hosting, security and DevOps frictions remained, and Modelence's diagnosis was that the issue was the connections between services rather than any individual one.
A second Hacker News launch followed in late January 2026, and discussion centered on TypeScript guardrails that raise the success rate of AI-generated code and on giving agents first-class auth, cron and deployment primitives. The open question is whether developers and enterprises will trust one young vendor's full stack over the best-in-class services they already use.
What has to be true
- Problem placement: Modelence attacked the seam where AI-generated apps actually fail — integration between services — instead of competing on raw code generation quality.
- Audience-first launch: releasing open source on Hacker News in August 2025 built developer credibility and feedback loops before the company had any funded marketing.
- Product as argument: one TypeScript+MongoDB stack with typed schemas directly addresses the class of errors AI code generators make, making 'it just ships' the demo rather than a promise.
- Validation: Y Combinator leading the $3M seed showed investors saw the plumbing layer, not the model layer, as the bottleneck for AI-built software.
What can be applied
Modelence sold the pain AI coders feel at the seams, not a better database: one pre-connected stack, it argued, is what lets generated code cross into production.
Aftermath
As of early February 2026, Modelence is live and seed-funded: the framework remains open source on GitHub, the AI app builder is generally available, and the team is marketing an 'agent-native cloud' story — one coherent stack where human developers and AI agents share the same TypeScript contracts. Its main challenge is incumbency: Google, Amazon and startups like Shuttle are chasing the same infrastructure opportunity, and Modelence must convince developers to bet their production apps on a young vendor's full-stack bet.
Sources
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card