Kimi/kimi-k2.5 cannot be called to execute tool calls in cursor

[yuanxs21]

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When I used the customed api (DashScope of Alibaba Cloud), I tried to employ the third-party api service from the Moonshot of the kimi/kimi-k2.5. However, the cursor raised an error after outputting a few words and calling reading tools a few times, as follows:

Provider returned error: {“error”:{“message”:“thinking is enabled but reasoning_content is missing in assistant tool call message at index 3”,“type”:“InvalidParameter”,“param”:null,“code”:“InvalidParameter”}}

I also tried other third-party models, such as MiniMax/MiniMax-2.5 and MiniMax/MiniMax-2.7 of the DashScope api, these models can run correctly.

Steps to Reproduce

1. Deploy customed api in the cursor editor.
2. Add kimi/kimi-k2.5 model.
3. Select kimi/kimi-k2.5 and talk with the model in the chat.

Operating System

MacOS

Version Information

Version: 2.6.21
VSCode Version: 1.105.1
Commit: fea2f546c979a0a4ad1deab23552a43568807590
Date: 2026-03-21T22:09:10.098Z
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 24.6.0

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

[deanrie]

Hey, thanks for the detailed report.

The error is coming from the DashScope API. When thinking mode is enabled for the kimi-k2.5 model, it requires the reasoning_content field in assistant messages that include tool calls. Right now, Cursor doesn’t preserve that field when it sends the conversation history back to the provider during multi-turn tool-call flows, so the request gets rejected.

This is a known limitation for thinking or reasoning models when using custom API keys. I’ve shared this with the team, but there’s no ETA for a fix yet.

As a temporary workaround:

• Use models without thinking mode through DashScope (for example, the MiniMax models you mentioned work fine)
• Or check if DashScope has an option to disable thinking mode for kimi-k2.5, which may fully avoid the issue

If you have any questions, let me know.