Vellum

Vellum is a freemium AI development platform in the LLM ops category that enables engineering and product teams to build, evaluate, and deploy production-grade AI applications, with a free Develop tier for prototyping, a Scale tier starting around $500 per month for production workloads, and custom Enterprise pricing for compliance-driven organizations.

The platform provides a collaborative prompt engineering environment where teams can version, test, and optimize prompts across multiple LLM providers—including OpenAI, Anthropic, Google, Cohere, and open-source models—without changing application code.

Founded in 2022 and headquartered in San Francisco, Vellum has grown to serve hundreds of companies ranging from startups to enterprises that rely on its infrastructure to move LLM features from prototype to production. The platform processes millions of LLM evaluations monthly and supports teams across industries including fintech, healthcare, legal tech, and e-commerce. As of early 2026, the company has over 60 employees and has raised more than $20 million in venture funding.

Vellum's core capabilities span three pillars: Build, Evaluate, and Deploy. The Build layer includes a visual workflow editor for designing complex LLM pipelines with branching logic, tool use, retrieval-augmented generation (RAG), and multi-step chains—all without writing boilerplate orchestration code. The Evaluate layer provides quantitative and qualitative testing frameworks, enabling teams to run automated regression tests on prompt changes, compare model outputs side-by-side, and track quality metrics over time using custom scoring functions or LLM-as-judge evaluators. The Deploy layer offers versioned API endpoints, A/B testing for prompt variants, real-time monitoring dashboards, and rollback capabilities so teams can ship with confidence.

Key differentiators include Vellum's model-agnostic architecture, which avoids vendor lock-in by letting teams swap LLM providers at the configuration level; its robust document processing and RAG pipeline tools for ingesting, chunking, and searching enterprise knowledge bases; and its emphasis on collaboration through shared workspaces, approval workflows, and audit trails designed for cross-functional teams. The platform also provides semantic search indexes, a prompt template registry, and detailed cost and latency analytics to help teams optimize both quality and spend.

Vellum supports over 50 LLM models, offers SOC 2 Type II compliance, and provides enterprise-grade features including SSO, role-based access control, and dedicated infrastructure options. The platform continues to expand its evaluation and observability tooling to meet growing demand for reliable AI application development.