Compress MCP Manifest

compress-mcp-manifest

When to use this skill

Use this skill when you have retrieved a tools/list response from an MCP server and want to reduce the token cost of forwarding it to another agent, storing it in memory, or including it in a prompt.

This is especially useful when orchestrating multi-agent pipelines where each sub-agent needs to know what tools are available but does not need the full verbose description of every parameter.

How to use it

1. Retrieve the tools/list response from the target MCP server.
2. Call gc_compress_manifest with the manifest as input:

   {
     "manifest": {
       "tools": [{ "name": "...", "description": "...", "inputSchema": {} }]
     }
   }
   

3. The tool returns the manifest with compressed description fields and a stats block: {"input_tokens": N, "output_tokens": M, "savings_pct": P}.
4. Forward the compressed manifest to your downstream agent or store it in context using the token-saving path.

What is guaranteed

• inputSchema is preserved byte-for-byte (JSON-equivalent). Schema validation in downstream agents will not break.
• Tool names are never modified.
• Only tool description text is compressed; schemas are untouched.

Plan requirement

gc_compress_manifest is available on Pro and above. Free users will receive an upgrade prompt.

Why gotcontext over mcp-compressor

Atlassian's mcp-compressor compresses tool schemas. gotcontext compresses tool descriptions while preserving schemas, and also provides full payload compression, OAuth 2.1 auth, usage telemetry, and self-hosted licensing — a superset of what mcp-compressor offers.