With tool definitions deferred, what does a connected MCP server still put into context at the start of every session?

Easyclaude-codemcpcontexttool-search

Last reviewed 21 Aug 2026

Its tool names and its server instructions, and nothing beyond those two.Correct answer
Nothing whatsoever, until one of the server's tools is called for the first time.
Only the tool names, since instructions are fetched later alongside the schemas.
Nothing visible, because names are withheld deliberately so the model cannot see them.

Show answer & explanation

Correct answer: A. Its tool names and its server instructions, and nothing beyond those two.

Explanation

The principle — Deferral trades detail for presence. Claude needs to know a server exists and roughly what it is for; it does not need every parameter until it uses one.

Why the key is correct — Only tool names and server instructions load at session start, so adding more MCP servers has minimal impact on your context window. Those two are the cost you pay per connected server.

Why the others are wrong — The cost is not zero. Instructions load with the names rather than later. And nothing is being hidden from the model; the names are precisely what it keeps.

Remember this — Names and instructions at the start, schemas on demand.

Sources — Anthropic's Claude Code MCP reference.

Sources

Primary sourceConnect Claude Code to tools via MCP — Anthropic · accessed 20 Aug 2026
“Only tool names and server instructions load at session start, so adding more MCP servers has minimal impact on your context window.”

Practise 10 questions on this topic

Take MCP Integration in Claude Code — Timed Test 1 (10 questions) — scored instantly, explanation for every question, no login.

Start the test

With tool definitions deferred, what does a connected MCP server still put into context at the start of every session?

Related questions

With tool definitions deferred, what does a connected MCP server still put into context at the start of every session?

Related questions