What happened?
It's able to use the tool and what I see in the terminal is accurate but according to Qwen, it's not able to see the result. I've only tested with the Figma MCP server I found that it's able to use the tool and the result of the tool call is visible and accurate in the terminal but the LLM or the Qwen Coder CLI is not able to actually see the result.
It fails silently and goes on performing the assigned task as if nothing happened. It was only after explicitly asking Qwen what was in the Figma mockup that I found it was just making things up, and after inquiring even further that I found it had no idea what was in the mockup. I'm using the https://openrouter.ai/api/v1 Base URL and anthropic/claude-sonnet-4.5 model. Using Claude with the same model and the anthropic base url works though I haven't tested claude with the openrouter endpoint.
│ data-node-id="38:521"> │
│
│
│
│
│
│ │
│ │
│ │
│ ); │
│ } │
│ SUPER CRITICAL: The generated React+Tailwind code MUST be converted to match the target │
│ project's technology stack and styling system. │
│ 1. Analyze the target codebase to identify: technology stack, styling approach, │
│ component patterns, and design tokens │
│ 2. Convert React syntax to the target framework/library │
│ 3. Transform all Tailwind classes to the target styling system while preserving exact │
│ visual design │
│ 4. Follow the project's existing patterns and conventions │
│ DO NOT install any Tailwind as a dependency unless the user instructs you to do so. │
│ │
│ Node ids have been added to the code as data attributes, e.g. data-node-id="1:2". │
│ Images and SVGs will be stored as constants, e.g. const image = │
│ These │
│ constants will be used in the code as the source for the image, ex: <img src={image} │
│ />. Image assets are stored on a remote server for 7 days and can be fetched using the │
│ provided URLs until they expire. │
│ IMPORTANT: After you call this tool, you MUST call get_screenshot to get a screenshot │
│ of the node for context. │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The tools are executing successfully but I'm not seeing the actual content in the response. This might be
a display issue on my end. Let me try a different approach - can you share a screenshot of the Figma
mockup or describe what you see? That would be the fastest way for me to understand exactly what fields
and layout the design specifies.
Alternatively, what specific differences are you noticing between what I built and what the mockup
shows? That will help me understand what needs to be changed.
/mcp
Configured MCP servers:
🟢 figma - Ready (10 tools, 1 prompt)
Tools:
- add_code_connect_map
- create_design_system_rules
- generate_diagram
- get_code_connect_map
- get_design_context
- get_figjam
- get_metadata
- get_screenshot
- get_variable_defs
- whoami
Prompts:
- create_design_system_rules
What did you expect to happen?
Qwen should be able to see the returned data of the tool call and use it accordingly.
Client information
Client Information
╭──────────────────────────────────────────────────────────────────────────╮
│ │
│ About Qwen Code │
│ │
│ CLI Version 0.6.2 │
│ Git Commit 40959d6 │
│ Model anthropic/claude-sonnet-4.5 │
│ Sandbox no sandbox │
│ OS Platform linux │
│ OS Arch x64 │
│ OS Release 6.12.65-1-lts │
│ Node.js Version v25.2.1 │
│ NPM Version 11.7.0 │
│ Session ID a1f893d9-ea04-4b0e-9fd3-0e27e2h37aje │
│ Auth Method openai │
│ Base URL https://openrouter.ai/api/v1 │
│ Memory Usage 294.6 MB │
│ │
╰──────────────────────────────────────────────────────────────────────────╯
Login information
API KEY, Base URL: https://openrouter.ai/api/v1, Model: anthropic/claude-sonnet-4.5.
This also occurred with the qwen-vl model and otherwise identical setup.
Anything else we need to know?
No response
What happened?
It's able to use the tool and what I see in the terminal is accurate but according to Qwen, it's not able to see the result. I've only tested with the Figma MCP server I found that it's able to use the tool and the result of the tool call is visible and accurate in the terminal but the LLM or the Qwen Coder CLI is not able to actually see the result.
It fails silently and goes on performing the assigned task as if nothing happened. It was only after explicitly asking Qwen what was in the Figma mockup that I found it was just making things up, and after inquiring even further that I found it had no idea what was in the mockup. I'm using the https://openrouter.ai/api/v1 Base URL and anthropic/claude-sonnet-4.5 model. Using Claude with the same model and the anthropic base url works though I haven't tested claude with the openrouter endpoint.
What did you expect to happen?
Qwen should be able to see the returned data of the tool call and use it accordingly.
Client information
Client Information
╭──────────────────────────────────────────────────────────────────────────╮
│ │
│ About Qwen Code │
│ │
│ CLI Version 0.6.2 │
│ Git Commit 40959d6 │
│ Model anthropic/claude-sonnet-4.5 │
│ Sandbox no sandbox │
│ OS Platform linux │
│ OS Arch x64 │
│ OS Release 6.12.65-1-lts │
│ Node.js Version v25.2.1 │
│ NPM Version 11.7.0 │
│ Session ID a1f893d9-ea04-4b0e-9fd3-0e27e2h37aje │
│ Auth Method openai │
│ Base URL https://openrouter.ai/api/v1 │
│ Memory Usage 294.6 MB │
│ │
╰──────────────────────────────────────────────────────────────────────────╯
Login information
API KEY, Base URL: https://openrouter.ai/api/v1, Model: anthropic/claude-sonnet-4.5.
This also occurred with the qwen-vl model and otherwise identical setup.
Anything else we need to know?
No response