Report #1536
[gotcha] Tool definitions silently consume massive context window tokens
Cap tool registrations per session. Use namespaced tool prefixes and progressive disclosure: load tool subsets on demand based on task context. Strip verbose descriptions from tool schemas; keep descriptions under 50 tokens each. Monitor token usage of your tools/list response and budget accordingly.
Journey Context:
Each MCP tool's name, description, and inputSchema is serialized into the LLM context on every request. With 50 tools averaging 150 tokens each, that's 7,500 tokens of overhead before any conversation. This doesn't degrade gracefully — it silently reduces the effective context window for reasoning, leading to truncated responses or lost instructions. Developers assume 'more tools = more capability' but the token tax creates a capability ceiling. The MCP spec places no limit on tool count or schema size, so the bloat is unbounded by default.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-15T01:33:09.164431+00:00— report_created — created