Use the Adobe Workfront MCP server
The information on this page refers to functionality not yet generally available. It is available only in the Preview Sandbox environment.
The Adobe Workfront MCP server lets you find, create, update, and manage Workfront items by asking an AI agentic platform in natural language. The platform decides which Workfront actions to call and handles the conversation with Workfront for you.
Prerequisites
- You must set up the connection between your AI agentic platform and the Workfront MCP server. For setup instructions, see Configure the Adobe Workfront MCP server.
- Your Workfront instance must be enabled on Adobe Identity Management System (IMS).
- You must have a Workfront account with the necessary access level and object permissions for the items you want to work with.
This article assumes you’ve already set up the connection. For information about setup, see Configure the Adobe Workfront MCP server.
Available tools
The Workfront MCP server exposes a set of tools that the AI agentic platform calls on your behalf. For example, tools to search Workfront, create items, update fields, and manage approvals. For the full reference list, see Adobe Workfront MCP server tools.
Examples of what to ask
After you’re connected, type natural-language requests in your AI agentic platform. The AI agentic platform decides which Workfront actions to call and returns the results.
Find and view your work
To search Workfront for items that match what you’re looking for, ask:
- Show me all active projects for the Brand Marketing team.
- Get me all tasks assigned to Joan Harris.
- Show me all issues in the “Website Redesign” project under the “Technical” category.
Create new items
To create projects, tasks, or other Workfront items, ask:
- Create a blank project called “Q2 Innovation Sandbox” starting March 10 and ending April 30. Set me as the owner.
- Add a new task called “Landing Page QA” to the project and schedule it from April 22 to April 26.
- Create a new campaign record called “Summer Sale 2026.”
Update existing items
To make changes to items already in Workfront, ask:
- Update the “Design Review” task so it finishes on April 18, and assign it to the creative team.
- For the “Lucent AI Launch – NA” project, push the finish to mid-April and increase the budget to $150K.
- Update the budget field in the “Summer Campaign” record to $75,000.
Delete items
To remove Workfront items, ask:
- Delete the project called “Q1 Test Campaign.”
- Remove the “Print Asset Production” task from the project.
- Delete the campaign record named “Old Promo.”
Work with approvals
To manage document and asset approvals, ask:
- Add Sarah Chen and Miguel Alvarez as approvers on the current document.
- Send a reminder to approvers on the asset “Spring Campaign Video” who haven’t responded.
- Apply the “Marketing Launch” approval template to the asset “Spring Campaign Video.”
Work with Planning records
To manage planning records, ask:
- Create a new planning record called “Q2 Marketing Plan” for the Brand Marketing team.
- Add a new task called “Social Media Audit” to the planning record.
- Update the “Social Media Audit” task so it finishes on April 18, and assign it to the creative team.
Work with workflow
To manage workflow, ask:
- Route the “Q2 Innovation Sandbox” project to the Innovation Review Board.
- Update the “Summer Campaign” record to the “Ready for Launch” status.
- Approve the “Summer Campaign” record.
Chain requests across a conversation
You can chain requests in a single conversation. The AI agentic platform keeps the context, so each request can build on the previous one. For example:
- Ask for a set of items: Find my overdue tasks.
- After you get the list, ask for an action on the results: Update all of them to next Friday.
Considerations
Keep the following considerations in mind when you use the Workfront MCP server:
The AI agentic platform may use information from earlier in the conversation
AI agentic platforms sometimes reuse data from earlier in a conversation instead of asking Workfront for the latest. If something has changed in Workfront since the AI agentic platform last looked, you might see outdated information.
To force the AI agentic platform to fetch fresh data, ask for it explicitly. For example:
- Get the latest data from Workfront. Don’t use cached results.
Check for updates to the Workfront MCP server
You might want to periodically refresh your connection to the Workfront MCP server to ensure that you have the latest tools and capabilities.
Most updates should happen automatically. However, we recommend checking the Workfront release notes periodically.
Data and security
The Workfront MCP server tools are consistent with how API calls work. Workfront doesn’t store prompts, responses, or any other data. Whatever Workfront data you ask for is accessible in the Workfront platform.
Your access level and object permissions determine what you can query for or write to Workfront. Your Workfront administrator has control of MCP read and write actions in the Workfront Setup area.
What data leaves Workfront
The AI agentic platform provider has access to the Workfront data you interact with through the Workfront MCP server. Check with your AI agentic platform provider for details.
How AI agentic platform providers handle your Workfront data
Workfront does not have control over how the AI agentic platform provider handles your Workfront data. Check with your AI agentic platform provider for details.
Troubleshoot day-to-day use
| table 0-row-3 1-row-3 2-row-3 3-row-3 | ||
|---|---|---|
| Problem | Likely cause | Fix |
| The AI agentic platform is giving you outdated information. | The AI agentic platform is reusing data from earlier in the conversation. | Ask for fresh data from Workfront. |
| The AI agentic platform returned data from the wrong Workfront items. | The AI agentic platform picked the wrong items based on ambiguous wording. | Ask again with more specific names, IDs, or filters. |
| An update or delete didn’t take effect in Workfront. | Your Workfront administrator has disabled write actions for the Workfront MCP server, or you don’t have permission to perform the action on the specific item. | Confirm with the AI agentic platform that the action ran. Then check that write actions are enabled for the Workfront MCP server and that you have permission to change the item. |
For more information about setup and authentication issues, see Troubleshoot setup and authentication in Configure the Adobe Workfront MCP server.
Frequently asked questions
What is an AI agentic platform?
An AI agentic platform is an AI tool that can take actions on your behalf in
other systems, not just answer questions. When you connect one to Workfront
through the MCP server, it can find, create, update, and delete Workfront
items based on what you ask it in natural language. Examples include Claude
Desktop, ChatGPT, and other MCP-compatible AI clients.
Do I need to be a Workfront administrator to use the Workfront MCP server?
No. Any Workfront user can use the Workfront MCP server through a connected
AI agentic platform. The AI agentic platform acts using your Workfront
account, access level, and object permissions, so you can only do what you
could already do directly in Workfront.
Why can’t the AI agentic platform create, update, or delete items for me?
Your Workfront administrator controls which MCP actions are allowed in the
Workfront Setup area. If write actions are disabled, the AI agentic platform
can still find and read Workfront items but can’t make changes. You also
need the right access level and object permissions for the specific items
you’re working with.
Will the AI agentic platform ask me before it changes or deletes Workfront data?
That depends on the AI agentic platform, not on Workfront. Most platforms
prompt you to confirm before they run an action, especially for deletes.
Before you approve a request, read what the platform says it’s about to do —
the changes happen in Workfront the same way they would if you made them
yourself in the interface.
Why did the AI agentic platform return the wrong Workfront items?
The AI agentic platform picks items based on the words you used. If your
request is ambiguous–for example, two projects have similar names–it
might pick the wrong one. Ask again with more specific names, IDs, dates,
or filters to narrow the results.
What Workfront items can I work with through an AI agentic platform?
Any items you have access to in Workfront through your access level and
object permissions. This includes projects, tasks, issues, documents,
approvals, Planning records, and more.
Can other people see my conversations with the AI agentic platform?
Workfront doesn’t store your prompts or the AI agentic platform’s responses.
Whoever provides your AI agentic platform controls how your conversations
are stored or shared. Check with your AI agentic platform provider for
details.
Do I need to know the Workfront API or which MCP tool to use?
No. The AI agentic platform translates your natural-language request into
the right Workfront actions and picks the right tools for you. If you’re
already familiar with the Workfront API, the actions will feel familiar,
but it isn’t a requirement.
Is my Workfront data sent to the AI agentic platform provider or stored by them?
For more information, see Data and security in this
article.
What happens when a new version of the Workfront MCP server releases?
The MCP server generally updates automatically, but you may need to refresh your connection to the MCP server at times to see the latest tools and capabilities.
Can I use the Workfront MCP server if my Workfront instance isn’t enabled on Adobe Identity Management System (IMS)?
No. Your Workfront instance must be enabled on Adobe Identity Management System (IMS) to use the Workfront MCP server. If you’re not sure whether your instance is enabled on IMS, contact your Workfront administrator.