AI Deck

Vellum — An AI agent building platform that unifies prompt management, evaluation, and monitoring

An AI agent and LLM application development platform offered since 2023 by Vellum AI, based in the United States. It brings prompt management and version control, workflow orchestration, evaluation (evals), RAG, deployment, and production monitoring together in a single environment. Both a visual workflow builder and Python/Go SDKs are available, so non-engineers and engineers can work from the same definition. It connects with more than 1,000 tools such as Slack, Notion, and HubSpot, and is used across industries from healthcare to finance. In recent years the company has also been developing Vellum Assistant, an open-source personal AI assistant, alongside the platform.

Key Features

  • Prompt management and versioning: Manage prompts in one place with change history, comparison, and rollback. You can validate the same prompt while swapping between multiple LLM providers
  • Visual workflow builder: Assemble an agent’s processing flow by connecting nodes for LLM calls, branching, tool execution, and more. The no-code UI and the code definition describe the same thing, so teams are less likely to drift apart in their understanding
  • Built-in evaluation and regression testing: Run a workflow in bulk against a set of test cases and compare accuracy and output quality numerically. Degradation caused by a prompt change can be caught before it reaches production
  • RAG (retrieval-augmented generation): Document indexing and semantic search are included, so you can build agents that reference your own data
  • Production monitoring and traces: Follow the behavior of running agents from node-level traces up to workflow-wide dashboards. Scheduled triggers for recurring runs are also supported
  • Integrations with 1,000+ tools: Connect to business tools such as Slack, Notion, and HubSpot to embed agents into internal processes
  • Flexible deployment: In addition to cloud usage, higher plans support running in a VPC or on-premises

Pricing

Vellum has separate pricing structures for the development platform and the personal assistant.

Vellum Platform (for development)

PlanMonthly priceWhat’s included
Free$01 user, a small allowance of credits for trials
Pro$251 user, expanded builder credits
Business$79 per userFor small teams, with a larger credit allowance
EnterpriseContact salesUnlimited credits, RBAC/SSO/environment separation, VPC install, dedicated support

Vellum Assistant (hosting for the personal assistant)

PlanMonthly priceWhat’s included
Free credits$0Your first tasks are free
Mighty$301 vCPU / 2 GiB RAM, 10 GB storage
Super$1002.5 vCPU / 5 GiB RAM, 30 GB storage, custom subdomain
Ultra$2004 vCPU / 8 GiB RAM, 60 GB storage, custom subdomain

Execution of AI responses, web searches, image generation, and similar work is billed as usage on a “$1 = 1 credit” basis. For the assistant, self-hosting the open-source version is also an option.

Pricing is as of August 2026. Plan structures change from time to time, so please check the official site for the latest information.

Pros & Cons

Pros

  • Prompt management, workflows, evaluation, and monitoring are consolidated, so you avoid stitching several tools together
  • Evaluation and regression testing are included by default, so quality regressions from prompt changes surface before release
  • The no-code UI and the SDK share the same definition, making it easier to divide work between non-engineers and engineers
  • VPC and on-premises options make it viable for industries with constraints on moving data outside their environment
  • A free plan is available, so you can start small and expand later

⚠️ Cons

  • Credit-based usage billing is layered on top, so costs are hard to forecast until your usage volume settles
  • Because the feature set is broad, it can feel heavy if you only need one thing, such as prompt management or monitoring alone
  • The Business plan’s per-user price is on the higher side, so costs grow quickly as the team expands
  • Having two product lines ─ the development platform and the personal assistant ─ can be confusing at first glance
  • Japanese-language official material and case studies are still limited, and documentation is primarily in English

Comparison with Similar Services

ItemVellumLangSmithDifyn8n
ProviderVellum AILangChainLangGeniusn8n GmbH
Main useBuilding, evaluating, and operating agentsMonitoring and evaluating LLM appsBuilding LLM apps and agentsGeneral-purpose workflow automation
Visual UIYes (workflow builder)LimitedYesYes
EvaluationBuilt inCore strengthYesNot included (LLM evals out of scope)
Self-hostingVPC/on-prem on higher plansAvailable on paid plansYes, via the open-source editionYes, via the open-source edition
Free to tryFree plan availableFree tier availableFree (self-hosted)Free (self-hosted)

Who Is It For

  • Teams whose prompts are scattered across spreadsheets and code, and who want to consolidate management in one place
  • Developers who want to evaluate LLM output quality with numbers rather than intuition, and to compare before and after changes systematically
  • Organizations that want business staff to assemble workflows and engineers to finish them
  • Practitioners who want to run RAG agents over internal data in production, with monitoring included
  • Companies with data-handling constraints that need to run in a VPC or on-premises

Summary

Vellum is a platform that consolidates the elements you need to run an LLM application in production ─ prompt management, workflows, evaluation, RAG, and monitoring ─ into a single product. Dedicated tools exist for each individual capability, but not having to stitch them together, plus having evaluation and regression testing built in from the start, is what usually makes Vellum worth choosing. Building one workflow on the free plan and running it through the evaluation features is a good way to judge whether it fits your development process.

← Blog