Fix
MCP Tool Not Working in an AI Coding Workflow
When an MCP tool is not working, check config path, auth state, tool approval, transport, logs and whether the agent is allowed to call it. Most failures are setup or permission issues, not model issues.
What should you check first?
Interactive diagram. Use Tab to move through hotspots or use the step controls when shown.
The task boundary was too broad or the prompt lacked non-goals.
Stop the run, save useful hunks and restart with one file group.
Pick the symptom, then use the smallest fix that can prove or disprove the cause.
- 1Stop the run and save the current diff.
- 2Check whether the task was too broad or missing files.
- 3Read the failing command, log or review comment out loud.
- 4Start a smaller run with one clear done state.
- 5Run the narrowest check before asking for more changes.
What usually caused it?
The agent did not see the file, error or rule that mattered.
The prompt asked for a result without naming the boundary.
The workflow lacked a test, typecheck or review gate.
Frequently asked questions
Who is MCP Tool Not Working in an AI Coding Workflow for?
Developers and platform teams debugging MCP in coding-agent workflows.
What makes this page credible?
The page starts with config, auth, approval, transport and logs.
What should I do next?
Start with one real repo task, capture the prompt and review the result before scaling the workflow.
Sources & last verified
Cursor ships frequently. Facts verified against primary sources on June 23, 2026.