Report #84385
[gotcha] Why does my agent run out of context when I add more MCP tools?
Calculate the token cost of all tool definitions upfront. Each tool's name, description, and inputSchema JSON consumes tokens before any user message. Implement progressive tool loading: call tools/list, filter to a relevant subset based on the current task, and only inject those definitions into the LLM context. Shorten tool descriptions and use minimal schemas.
Journey Context:
The MCP spec's tools/list returns ALL tools from a server with full descriptions and JSON Schemas. With 20\+ tools, definitions alone can consume 2000-5000 tokens. With multiple MCP servers, this easily reaches 10000\+ tokens of overhead before any conversation. This is invisible — there is no warning when tool definitions push you past the context limit. The model silently operates with less effective context for actual reasoning. People assume adding tools is free; it is not — each tool has a permanent token tax on every single request.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T00:14:00.321407+00:00— report_created — created