Enable an MCP server because the current task needs its live data—not because it might be useful someday.
The decision behind the technique
Every enabled server can add tool definitions and supporting context before useful work begins. The cost is easy to miss because it happens before your visible prompt is answered.
Enable an MCP server because the current task needs its live data—not because it might be useful someday.
A practical step-by-step workflow
Use this sequence as a default, then tighten or expand it only when the evidence changes.
- List the external systems this task actually needs
- Disable unrelated servers before starting the task
- Re-enable a server only when the next step requires it
- Compare usage on one representative task
Failure boundary and finish line
Do not remove a server in the middle of a workflow that still depends on its tool results.
Finish when the configuration is smaller, reproducible and still preserves every required safety boundary.
Frequently asked questions
What should I do first?+
List the external systems this task actually needs
What is the main failure mode?+
Do not remove a server in the middle of a workflow that still depends on its tool results.
OFFICIAL SOURCES
Verify against the source
OpenAI Codex pricing and usageOpenAI MCP guidanceThis guide provides general information, not an OpenAI allowance promise or support decision. Your account and official pages govern actual usage, benefits and recovery times.