Understanding Bynder MCP

Summary

AI assistants are becoming a standard part of how teams work, but they only know what you tell them. The Model Context Protocol (MCP) is an open standard that lets AI tools connect directly to the systems your business already uses. Instead of copying and pasting information into a chat window, the assistant can look things up for itself.

Bynder MCP is that connection for your DAM. Once set up, your AI assistant can search for approved assets and get answers from brand guidelines directly inside tools like Claude, Microsoft Copilot, ChatGPT, and Gemini. No tab switching, no copying asset links, no maintaining a separate copy of your guidelines.

Who?

Any organization with an active Bynder DAM portal can use Bynder MCP for asset search and retrieval. To use brand guidelines through the MCP, a CX User Community (CXUC) package is also required.

Users connect with their own Bynder credentials, so no additional permissions need to be configured. If you can access an asset in Bynder, you can find it through the MCP.

Don't yet have Bynder? Start Here!

Why?

Find approved assets without leaving your AI tool. Ask for the asset you need in plain language and the MCP searches your DAM on your behalf, using your portal's existing taxonomy, metadata, and tags to surface the right file.

Users with an AI Search enabled portal can also leverage Natural Language Search to query assets without relying on their metadata and taxonomy. 

Get reliable brand guidance on the spot. Ask about tone, voice, colors, fonts, logos, or campaign-specific rules and get an answer drawn directly from your published CX User Community content. Answers always reflect the current published version, with no manual re-sync needed.

Your governance stays intact. Bynder’s MCP server applies the same access permissions already present in the DAM. No content is exposed that the user could not already see in the web app.

Example queries

  • "Find the latest approved hero image for the summer campaign."
  • "What is our primary brand color?"
  • "What tone of voice should I use for external communications?"

How?

Setup involves creating an OAuth2 App in your Bynder portal and connecting it to your AI client of choice. Full step-by-step instructions, including connection details for Claude, Copilot, ChatGPT, and Gemini, are available in the Bynder MCP technical documentation.

Tips for better results

The quality of search results reflects how well your DAM is structured. Portals with well-tagged assets, descriptive titles, and consistent metaproperties will get more accurate and relevant results from the MCP. Enrichment Agents from Bynder can help ensure a deeply enriched, easily searchable taxonomy, while Bynder’s AI Search package can help automatically tag assets using AI Facial Recognition and Text-In-Image Search that can be accessed via the MCP. 

FAQs

General FAQs

Do I need to set up anything in Bynder before connecting? Yes, an administrator needs to create an OAuth2 App in your Bynder portal first. This is a one-time setup. See the Bynder MCP technical documentation for full instructions.

Can all users connect, or does an admin need to do something first? An admin creates the OAuth2 App and shares the connection details with users. Individual users then connect using their own Bynder credentials in their AI client of choice.

What AI tools does Bynder MCP work with? Any AI client that supports MCP, including Claude, Microsoft Copilot, ChatGPT, and Gemini. See the technical documentation for client-specific setup guides.

I do not have CX User Community. Can I still use Bynder MCP? Yes, asset search and retrieval works without CX User Community. You will only see the brand guidelines tools if your organization has a CX User Community package.

Does this work alongside other MCPs and integrations? Yes. The Bynder MCP works alongside other MCP connections in your AI tool, such as a PIM, CMS, or marketing automation system, to support connected workflows.

Security and Access Control FAQs

Will the AI see assets I am not supposed to access?

No. By design, every MCP connection authenticates as the individual user via delegated OAuth. Each MCP tool call runs against the Bynder Public API using that user's own Bearer token. The same permission model (asset, collection, group, and guideline visibility) that governs the DAM portal today governs the MCP. The MCP tools only return what the connecting user's existing permissions already allow. We do not provide an elevated or separate access path.

Does each user connect under their own identity, or does the integration use a shared identity?

Users connect through delegated OAuth under their own identity. On the first connection, the user redirects to our OAuth/OIDC login and signs in with their existing DAM credentials. The resulting MCP session is scoped exactly to what that person can already see in the portal. We do not use a shared or service-level identity across users.

Do thumbnail and asset URLs require Bynder authentication, or could they be shared outside the platform?

They require authentication. MCP access requires OAuth authentication, and a valid session retrieves data according to that user's existing permissions. The only exception is assets a customer explicitly makes public, which are accessible without authentication just like any public asset today.

Why does the technical documentation list meta.assetbank:write access if the tools perform read or search operations?

The metaproperty tools (asset_search_metaproperties, asset_list_metaproperties, and asset_list_metaproperty_options) require the meta.assetbank:write scope, even though they only list and search. This is a technical requirement of the metaproperty functionality. It’s not a sign that the MCP can change anything in your portal.

Having the scope is not enough on its own. The scope only allows the MCP to ask for metaproperty data. The user who signs in must also have the correct permissions in your portal to access metaproperties. If they don’t, the metaproperty tools won’t be accessible to them, even if you included the scope when you created the OAuth app.

The other tools (asset search, asset download, tags, and CX User Community) don’t need this scope.

If you leave the scope out when creating your OAuth app, the MCP still works. Users can use everything except the metaproperty tools. You choose the scopes when you create the OAuth app, so you decide whether to include it.

How do I get an access token for the MCP?

First, create an OAuth2 App in your own Bynder portal. You own the app, and creating it gives you a Client ID and a Client Secret.

Then use that Client ID and Secret with our standard authorization, token and refresh endpoints. These return a short-lived access token in the form of a JSON Web Token (JWT).

Where are credentials stored, and how are they secured?

The access token we issue is a JWT. It is generated and digitally signed by Bynder, and it contains all the data needed to validate it. Because of this, we don’t store the tokens we issue.

Where a user signs in with a username and password, we never store those credentials in plaintext. We use one-way hashing (PBKDF2 with HMAC/SHA2) with an iteration count that is reviewed annually, currently set to 160,000.

Customers who use SSO authenticate through their own identity provider, so their login credentials are not stored by Bynder.

What MCP activity is captured in audit logs, and what customer data do we retain?

MCP tool calls pass the user's query text to the relevant MCP service and the Bynder Public API to retrieve matching metadata already held in your portal. We do not duplicate the query text and returned content into a separate store. Separately, we capture audit metadata for each MCP call, including request ID, tenant ID, the tool invoked, the resource IDs returned, and the calling user. These logs record metadata about the call, not the full query text or the returned content itself. They exist on our side today but are not currently exposed to customers. We are evaluating customer-facing log exports as a future capability.

Which models, infrastructure providers, or subprocessors support the MCP capability?

The MCP platform runs on our existing AWS infrastructure, fronted by Kong and an internal gateway service. We do not introduce any additional subprocessors for MCP. Enabling MCP does not change existing contractual terms, data-processing terms, or our subprocessor list.

Has the MCP capability undergone an independent security assessment?

We have completed an internal penetration test against the MCP implementation. This was informed by threat-modeling documentation and confirmed the correct implementation of the MCP features with no critical or high findings. MCP will also be included in our external penetration test scheduled for September/October 2026.

Related Articles

Updated