August 2025

MCP in Agent node

August 21st, 2025

We now support MCP as a tool in Agent node. Select MCP server from + Tool button within Agent node and connect to your remote MCP server.

Add MCP Server

Configure MCP Server

Agent node automatically handles tool discovery in the background. Once setup is complete, you’ll see all available tools displayed and ready to use.

MCP Server Tools

Workflow Sharing

August 21st, 2025

We have added the ability to share your Workflows with others. To share your Workflow, click the Share button on the top right of the workflow. You can toggle if you want the workflow to be publicly available for anyone with a link or you can share internally by keeping the Public Access toggle off.

Edit Code Preview Files Toggle

Editable Code Preview Files

August 21st, 2025

We now have the ability to edit files from the Code Preview. Before, you had to run vellum workflows pull to edit your code and then do a vellum workflows push to update the code of your workflow. Now you can click the toggle to enable Edit mode and modify the files straight from the Code Preview UI. Edit Code Preview Files Toggle

API Keys Removed from Side Nav

August 19th, 2025

We used to have an item in our side navigation called “API Keys.” Over time, this had become a dumping ground of miscellaneous api-related settings.

We’ve since found new, more reasonable, homes for everything that used to be on that page:

Improved Variable Detection When Pasting Prompts

August 15th, 2025

We’ve improved variable detection when pasting prompts into Rich Text inputs. Previously, when you pasted a prompt containing {{ variable }} placeholders that weren’t already defined in Vellum, these placeholders would be inserted as plain text rather than being recognized as variables.

Now, undefined variables are automatically created, and {{ variable }} placeholders will render as variable chips as expected. This should make it easy to copy/paste existing prompts from your codebase or other tools into Vellum.

Models Page Revamp

August 12th, 2025

We’ve launched a major overhaul of our Models Page to make it easier to discover and manage the AI models available in your Workspace. Previously, we had one large grid that listed all models in Vellum, which could be overwhelming and made it difficult to find specific models of interest. But no more!

Screenshot of the new Models Page showing a grid of model provider cards including Anthropic, AWS Bedrock, Azure AI Foundry, Azure OpenAI, BaseTen, and others with model counts and newest model information
New Models Page showing all model providers that Vellum integrates with

Key improvements include:

  • Provider-Based Navigation: When you click on “Models” in the side nav, you’ll now see all model providers that Vellum integrates with, organized as easy-to-browse cards
  • Provider-Level Configuration: Each provider page now includes provider-level settings, most commonly where you’ll enter your LLM Provider API Keys
  • Enhanced Model Discovery: Click into any provider to see all models from that specific provider with improved search, sort, and filter capabilities
Screenshot of the OpenAI provider page showing a list of GPT models with search functionality, status indicators, and provider configuration options including API key setup
Individual provider page showing OpenAI models with search, filtering, and provider configuration
  • Improved Organization: Models are now organized by provider, making it much easier to find models from specific providers like OpenAI, Anthropic, Google, and others
  • Better Search & Filtering: Enhanced search and filtering capabilities within each provider’s model list
  • Cleaner Interface: A more focused, less cluttered experience that scales better as we continue adding new model providers and models

This update makes it significantly easier to navigate Vellum’s extensive model catalog, which now includes 23 different model providers and hundreds of individual models. Whether you’re looking for the latest GPT models from OpenAI or exploring options from Anthropic, Google, or other providers, we hope this update makes it easier to find them!

Updated Node Handles UX

August 11th, 2025

We’ve updated the look and feel of Node Handles in the Workflow Builder to provide better visual feedback and improved usability:

Screenshot showing the updated Node Handles with different visual states for nodes with 0, 1, and 2+ incoming edges, plus new plus icons on source handles
Updated Node Handles showing different visual states for connection points

Key improvements include:

  • Visual States for Incoming Edges: Node handles now display different visual indicators based on the number of incoming connections:
    • Nodes with 0 incoming edges
    • Nodes with 1 incoming edge
    • Nodes with 2+ incoming edges
  • New Plus Icon: Source handles now feature a plus icon for clearer identification
  • Improved Edge Dropping: Enhanced interactions when connecting and dropping edges between nodes

These updates make it easier to understand your Workflow’s connection structure at a glance and provide a more intuitive experience when building complex workflows.

Video Input Support

August 8th, 2025

We’ve added support for video inputs to Prompts and Workflows when using supported models.

Composio Tool support

August 7th, 2025

We’ve added support to connect to over 250+ apps and 10K+ tools through our partner integration with Composio.

Composio makes it easy to connect with external data sources, APIs, and services, allowing you to extend the capabilities of your Vellum workflows. It’s a powerful way to integrate with tools like Google Sheets, Slack, Salesforce, and many more, enabling you to build complex workflows that interact with a wide range of external systems. It’s perfect for those who want to build internal tools or automate business processes.

Once all set up, you can add Composio as a Tool in Agent Nodes (formerly known as Tool Calling Nodes) in your workflows.

Add Composio Tool

You can then use Composio to call any of the supported tools in your Workflow. Composio Tool Selector

To make use of Composio, you’ll need to:

  1. Sign up for a Composio account
  2. Configure the integrations you’re interested in
  3. Create a Composio API Key
  4. Add the Composio API Key to as an Environment Variable in your Vellum Workspace

Insert Node Panel UX Improvements

August 7, 2025

We’ve updated the Insert Node Panel in the Workflow Builder with several UX improvements to make it easier to find and add nodes to your workflows:

Screenshot of the new Insert Node Panel showing the search bar, reorganized cards, and nested hierarchy with Basic Nodes and Execution Logic sections
Updated Insert Node Panel with search functionality and improved organization

Key improvements include:

  • Search Bar: Quickly find specific nodes by typing their name
  • Updated Copy: Clearer, more concise descriptions for each node type
  • Smaller Cards: More compact design that fits more nodes on screen
  • Reorganized Layout: Better categorization with logical groupings
  • Nested Hierarchy: Expandable sections for Control Flow and other categories
Screenshot showing the expanded Control Flow section with various node types including Guardrail, Error, Conditional, Map, and Merge nodes
Control Flow section with nested hierarchy showing Guardrail, Error, Conditional, Map, and Merge nodes
  • Split Subworkflow Options: Subworkflows are now clearly separated into “Start from Scratch” and “Deployed Subworkflow” options for better clarity
Screenshot of the Subworkflow section showing two distinct options: Start from Scratch and Deployed Subworkflow
Subworkflow section showing the split between Start from Scratch and Deployed Subworkflow options

We hope that these improvements make it easier to find the building blocks you need to create powerful AI workflows in Vellum. As always, we welcome your feedback on these changes!

GPT 5 on OpenAI

August 7, 2025

We’ve added support for OpenAI’s latest GPT-5 models. You can find both API variants on our Models page for each GPT-5 Model:

  • GPT-5 via Chat Completions API
  • GPT-5 via Responses API
  • GPT-5 Mini via Chat Completions API
  • GPT-5 Mini via Responses API
  • GPT-5 Nano via Chat Completions API
  • GPT-5 Nano via Responses API

GPT OSS 120B on Cerebras

August 5th, 2025

We’ve added support for GPT OSS 120B on Cerebras.

GPT OSS Models on Groq

August 5th, 2025

We’ve added support for GPT OSS 120B and 20B on Groq.

Anthropic’s Claude Opus 4.1

August 5th, 2025

We’ve added support for Anthropic’s latest model Claude Opus 4.1.

OpenAI’s GPT OSS Models

August 5th, 2025

We’ve added support for OpenAI’s two latest models:

Cerebras Qwen 3 480B Models

August 5th, 2025

We’ve added support for the three newest variants of the Qwen 3 480B models:

  • Qwen 3 Coder 480B on Cerebras
  • Qwen 3 Thinking 480B on Cerebras
  • Qwen 3 Instruct 480B on Cerebras

Vertex AI Gemini 2.5 flash lite for Finetuning

August 4th, 2025

We’ve added support for Gemini 2.5 Flash finetuned models. You can add finetuned Vertex AI models to your Workspace by selecting “Fine-tuned Vertex AI Models” on the models page.