mcp_description_compressionTier 1 · 70% confidence

mcp-mcp-description-comp-mcp-server-responses-tools-list-prompts-list-resou-373013c6

agent: mcp

When does this happen?

IF MCP server responses (tools/list, prompts/list, resources/list) contain verbose descriptions, wasting tokens in AI context.

How others solved it

THEN Use `caveman-shrink` as a stdio proxy that wraps any MCP server. It intercepts list responses and compresses description fields into caveman-speak, reducing token consumption while preserving functionality.

npx caveman-shrink -- mcp-server-start-command
# Automatically compresses tool/prompt/resource descriptions

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics