Skip to main content
Glama
glassflow

bloat

Official
by glassflow
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Remote

bloat-mcp

A deliberately bloated MCP server for a context-rot demo. Its one tool, search_handbook, returns the whole (synthetic) company handbook, about 44 KB (roughly 15k tokens), instead of the matching paragraph. Each call answers one question, so three questions cost three copies. Stateless, no auth, no storage, no logging.

Install (Claude Code)

claude mcp add --transport http bloat https://bloat-mcp.vercel.app/mcp

Related MCP server: Halyard Knowledge Base MCP Server

The prompt

Use the bloat handbook search to answer three questions: how many vacation days do full-time staff get, how soon must expense reports be filed, and how often are laptops replaced? Search once per question. Save the answers to answers.md, then read the file back and tell me what it says.

The file write and read-back matter: Rius's context rot rule fires when one tool's results take 30% or more of a model call's input and hold for the next 2 calls.

Run /context before and after to watch the context window jump, then look at the trace in your Rius console.

Deploy

vercel deploy --prod

Local test: node server.js 8787, then claude mcp add --transport http bloat http://127.0.0.1:8787/mcp.

Keep each result under about 50 KB: Claude Code spills larger MCP results to a file instead of inlining them, which defeats the demo.

MIT licensed.

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • Aleph

    Hmbown
    A
    license
    Not graded
    quality
    D
    maintenance
    Enables AI assistants to analyze documents larger than their context window by loading files into RAM and querying them via search, navigation, and Python execution tools. Supports recursive reasoning to process massive datasets in chunks using sub-agents.
    40 PyPI
    219
    MIT

  • ponylink-rules-mcp

    Sejin-Koo
    F
    license
    Not graded
    quality
    C
    maintenance
    Provides read-only semantic search over the company's 36 internal rule documents (e.g., employment, HR, disciplinary, attendance, salary regulations). It exposes tools to query the knowledge base by natural language and to retrieve total chunk count, with no modification capabilities.
    -

  • enterprise-knowledge-mcp

    Shruti-Gorhe
    F
    license
    Not graded
    quality
    B
    maintenance
    Enables natural-language retrieval from enterprise policy documents via RAG, exposing search_enterprise_knowledge and get_document_sources tools.
    1
    -