Hello everyone,
problem/question
I’ve encountered an interesting issue while working with MCP servers, and I’m hoping someone can explain what’s happening or suggest a fix.
When using mcp-atlassian ( GitHub - sooperset/mcp-atlassian: MCP server for Atlassian tools (Confluence, Jira) ) with the Qwen3 model, the AI agent fails to interact properly with the MCP server.
For example, when I ask the model “what can you do?”, it correctly detects that a tool is connected. However, when I request something like “show task ABC-123”, no request is actually sent to the MCP server — there’s no trace of any call in the logs.
Interestingly, when I connect another MCP server — GitHub - nguyenvanduocit/jira-mcp: A Go-based MCP (Model Control Protocol) connector for Jira that enables AI assistants like Claude to interact with Atlassian Jira. This tool provides a seamless interface for AI models to perform common Jira operations including issue management, sprint planning, and workflow transitions. - everything works as expected.
Additionally, when using ChatGPT, both MCP servers (mcp-atlassian and jira-mcp) work correctly.
Has anyone faced a similar issue or found a workaround for this behavior? Any insights would be greatly appreciated.
Thanks in advance!
Please share your workflow
Information on your n8n setup
- n8n version: 1.119.0
- **Database (default: SQLite):**SQLite
- n8n EXECUTIONS_PROCESS setting (default: own, main):
- Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
- Operating system: ubuntu 24.04.3
Information on your softvare setup
- inference server: ollama
- LLM: qwen3:4b-instruct-2507-fp16
- problem mcp server: GitHub - sooperset/mcp-atlassian: MCP server for Atlassian tools (Confluence, Jira)
- normal mcp server: GitHub - nguyenvanduocit/jira-mcp: A Go-based MCP (Model Control Protocol) connector for Jira that enables AI assistants like Claude to interact with Atlassian Jira. This tool provides a seamless interface for AI models to perform common Jira operations including issue management, sprint planning, and workflow transitions.