# Velvet

**Canonical:** https://apis.io/providers/velvet/  
**Website:** https://velvetlab.ai  
**APIs profiled:** 0

Velvet is a multimodal AI model R&D company (the "multimodal data lab") based in San Francisco and backed by Y Combinator's Fall 2025 batch. It partners with frontier AI labs and enterprises to build state-of-the-art video datasets and evaluation frameworks that improve models' spatial reasoning and interactive, low-latency capabilities. Velvet operates an invite-only platform for sourcing and contributing multimodal training data, with contributor payouts via Stripe and Wise. It was founded by Lucas Mantovani (CEO, formerly video data at Meta FAIR) and Lucas Tucker (CTO, formerly Adobe). As of this enrichment pass the company publishes no public API, developer documentation, or SDK.

## Kin Score — 13.5 / 100 (emerging)

Scored 2026-08-20 under rubric 0.12.0. Trend: flat (+0.0 from 13.5).

| Facet | Score |
|---|---|
| Discoverability | 57.4 |
| Contract Quality | 0.0 |
| Governance | 0.0 |
| Contract Governance | 0.0 |
| Operational Transparency | 0.0 |
| Developer Ergonomics | 4.8 |
| Commercial Clarity | 34.2 |
| Access Clarity | 34.2 |

## Agent readiness — 0.0 (human-only)

| Dimension | Value |
|---|---|
| Spec Presence | no |
| Agentic Access | no |
| Reversibility Documented | no |
| MCP Server | no |
| Auth Clarity | no |
| Idempotency | no |
| Error Semantics | no |
| OpenAPI Examples | no |
| Rate Limit Signal | no |
| Event Surface Described | no |
| Agent Skills | no |
| Well Known Catalog | no |
| Consent Identity | no |
| Agent Card | no |
| Dry Run Mode | no |

## Access

Requires approval — onboarding: approval, pricing: unknown, trial: no (confidence: medium).

## Security (1)

- **Velvet Domain Security** — TLSv1.3 · HSTS · DMARC

## Tags

Company, Artificial Intelligence, Multimodal AI, Machine-Learning, Datasets, Model Evaluation, Video, World Models, Y Combinator

---

Profiled by [API Evangelist](https://apievangelist.com) and published on [APIs.io](https://apis.io/providers/velvet/). Scores are computed from the provider's own public artifacts under a published rubric.