[GH-ISSUE #150] feat: Enhanced multi-chain tools invocation capabilities, refer to Cherry Studio for reference. #69

New issue

Closed

opened 2026-10-18 21:11:13 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-18 21:11:13 +00:00

Owner

Originally created by @AlbertNie on GitHub (May 20, 2025).
Original GitHub issue: https://github.com/open-webui/mcpo/issues/150

Check Existing Issues

I have searched the existing issues and discussions.

Problem Description

I have configured multiple MCP Servers. When users ask questions, they typically require sequential invocation of multiple MCP servers to obtain the final answer. However, the OpenWebUI + MCPo setup is underperforming. From observations, the large model stops after invoking MCPo twice and provides an incorrect answer. Additionally, the thinking and invocation processes are not displayed in real-time during the process. In contrast, Cherry Studio works differently: it allows the large model to first retrieve the tools list, then deliberate which interface to invoke first. After obtaining the result, it continues to deliberate which interface to invoke next based on the result, proceeding step-by-step until the correct answer is provided. All these invocation steps are executed in a continuous and transparent manner.

Desired Solution you'd like

I hope that OpenWebUI can achieve MCP capabilities equivalent to those of Cherry Studio.

Alternatives Considered

No response

Additional Context

No response

Originally created by @AlbertNie on GitHub (May 20, 2025). Original GitHub issue: https://github.com/open-webui/mcpo/issues/150 ### Check Existing Issues - [x] I have searched the existing issues and discussions. ### Problem Description I have configured multiple MCP Servers. When users ask questions, they typically require sequential invocation of multiple MCP servers to obtain the final answer. However, the OpenWebUI + MCPo setup is underperforming. From observations, the large model stops after invoking MCPo twice and provides an incorrect answer. Additionally, the thinking and invocation processes are not displayed in real-time during the process. In contrast, Cherry Studio works differently: it allows the large model to first retrieve the tools list, then deliberate which interface to invoke first. After obtaining the result, it continues to deliberate which interface to invoke next based on the result, proceeding step-by-step until the correct answer is provided. All these invocation steps are executed in a continuous and transparent manner. ### Desired Solution you'd like I hope that OpenWebUI can achieve MCP capabilities equivalent to those of Cherry Studio. ### Alternatives Considered _No response_ ### Additional Context _No response_

bsakaguchi closed this issue

2026-10-18 21:11:13 +00:00

bsakaguchi referenced this issue

2026-10-18 21:12:20 +00:00

[PR #69] [MERGED] Feat/sse mcp server connection #162

bsakaguchi referenced this issue

2026-10-18 21:12:21 +00:00

[PR #65] [MERGED] 0.0.11 #163