[GH-ISSUE #16883] feat: Expose OpenWebUI API Endpoints as an MCP Server (Callable Tools) #18084

New Issue

GiteaMirror · 2026-04-20T00:01:19-05:00

GiteaMirror commented

2026-04-20 00:01:19 -05:00

Originally created by @flefevre on GitHub (Aug 25, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16883

Check Existing Issues

I have searched the existing issues and discussions.

Problem Description

Summary

I propose adding support for exposing the existing OpenWebUI API endpoints through an MCP server implementation.
This would allow OpenWebUI to act as its own "callable tool provider," enabling models and assistants created within the platform to invoke OpenWebUI’s own functionality directly.

Motivation

Currently, OpenWebUI provides a great set of API endpoints (see documentation) for managing models, users, and knowledge bases. However, these endpoints are not yet integrated as callable tools in an MCP-compatible workflow.

By exposing them as MCP tools, OpenWebUI could:

Self-manage: administrators could use natural language to execute actions (e.g., "Create the group alpha and add the following users").
Assist users dynamically: end-users could query assistants with prompts like "Find the right knowledge base or model to address this topic", and the MCP server would return relevant OpenWebUI resources.

This bridges the gap between assistant configuration and platform administration in a seamless, conversational way.

Desired Solution you'd like

Proposal

Implement an official OpenWebUI MCP server that wraps around the REST API endpoints.
When creating or configuring a model assistant in the panel, allow selecting the OpenWebUI MCP server as a tool.
Each API endpoint would be callable as a structured MCP tool function.
Provide a starter config/example so users can easily enable this integration.

Example Use Cases

1. Administrator workflow

Prompt: "Create the group alpha and add Alice, Bob, and Charlie."
Action: MCP server maps this request to the groups and users endpoints.

2. User workflow

Prompt: "Find the best model for answering questions about medical terminology."
Action: MCP server queries the models or knowledge base endpoints and returns a matching assistant description.

Benefits

Unified ecosystem: OpenWebUI becomes both the platform and the tool provider.
Lower barrier for automation: no need for external scripting to call the API.
Extensible: any new OpenWebUI endpoint can be surfaced automatically as a tool.

Alternatives Considered

No response

Additional Context

No response

Originally created by @flefevre on GitHub (Aug 25, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16883 ### Check Existing Issues - [x] I have searched the existing issues and discussions. ### Problem Description ## Summary I propose adding support for exposing the existing **OpenWebUI API endpoints** through an **MCP server** implementation. This would allow OpenWebUI to act as its own "callable tool provider," enabling models and assistants created within the platform to invoke OpenWebUI’s own functionality directly. ## Motivation Currently, OpenWebUI provides a great set of API endpoints ([see documentation](https://openwebui.com/getting-started/api-endpoints)) for managing models, users, and knowledge bases. However, these endpoints are not yet integrated as callable tools in an MCP-compatible workflow. By exposing them as MCP tools, OpenWebUI could: - **Self-manage**: administrators could use natural language to execute actions (e.g., *"Create the group `alpha` and add the following users"*). - **Assist users dynamically**: end-users could query assistants with prompts like *"Find the right knowledge base or model to address this topic"*, and the MCP server would return relevant OpenWebUI resources. This bridges the gap between **assistant configuration** and **platform administration** in a seamless, conversational way. ### Desired Solution you'd like ## Proposal - Implement an **official OpenWebUI MCP server** that wraps around the REST API endpoints. - When creating or configuring a model assistant in the panel, allow selecting the **OpenWebUI MCP server** as a tool. - Each API endpoint would be callable as a structured MCP tool function. - Provide a starter config/example so users can easily enable this integration. ## Example Use Cases ### 1. Administrator workflow - **Prompt**: *"Create the group `alpha` and add Alice, Bob, and Charlie."* - **Action**: MCP server maps this request to the `groups` and `users` endpoints. ### 2. User workflow - **Prompt**: *"Find the best model for answering questions about medical terminology."* - **Action**: MCP server queries the `models` or `knowledge base` endpoints and returns a matching assistant description. ## Benefits - **Unified ecosystem**: OpenWebUI becomes both the **platform** and the **tool provider**. - **Lower barrier for automation**: no need for external scripting to call the API. - **Extensible**: any new OpenWebUI endpoint can be surfaced automatically as a tool. ### Alternatives Considered _No response_ ### Additional Context _No response_

GiteaMirror closed this issue

2026-04-20 00:01:20 -05:00

GiteaMirror referenced this issue

2026-04-20 00:43:16 -05:00

[GH-ISSUE #18084] issue: DEV docker image broken #18486

GiteaMirror referenced this issue

2026-04-25 07:54:32 -05:00

[GH-ISSUE #18084] issue: DEV docker image broken #34015

GiteaMirror referenced this issue

2026-05-05 20:40:03 -05:00

[GH-ISSUE #18084] issue: DEV docker image broken #57152

Sign in to join this conversation.