HTTP MCP server for research. Three tools, orientation-first, built for agents that run multi-pass research loops.

Built on mcp-use. No stdio, HTTP only.

Also exposes /health, health://status, and two optional MCP prompts: deep-research and reddit-sentiment.

Call start-research once at the beginning of each session with your goal. The server returns a brief that tells the agent exactly which tool to call first (reddit-first for sentiment/migration, web-first for spec/bug/pricing, both when opinion-heavy AND needs official sources), what keyword seeds to fire, and when to stop.

Pair the server with the run-research skill for the full agentic playbook:

npx -y skills add -y -g yigitkonur/skills-by-yigitkonur/skills/run-research

# from npm
HOST=127.0.0.1 PORT=3000 npx -y mcp-researchpowerpack

# from source
git clone https://github.com/yigitkonur/mcp-researchpowerpack.git
cd mcp-researchpowerpack
pnpm install && pnpm dev

Connect your client to http://localhost:3000/mcp:

{
  "mcpServers": {
    "research-powerpack": {
      "url": "http://localhost:3000/mcp"
    }
  }
}

Copy .env.example, set only what you need. Missing keys don't crash the server — they disable the affected capability with a clear error.

Any OpenAI-compatible endpoint. LLM_API_KEY, LLM_BASE_URL, and LLM_MODEL are all required together. Reasoning effort is always low.

pnpm test:evals writes a JSON artifact to test-results/eval-runs/<timestamp>.json.

When an OpenAI API key is present, it performs a live Responses API + remote MCP evaluation. Without an API key, it exits successfully in explicit skip mode and records that skip in the artifact.

Useful env vars:

• EVAL_MCP_URL
• EVAL_MODEL
• EVAL_API_KEY or OPENAI_API_KEY

pnpm install
pnpm dev          # watch mode, serves :3000/mcp
pnpm typecheck    # tsc --noEmit
pnpm test         # unit + http integration tests
pnpm build        # compile to dist/
pnpm inspect      # mcp-use inspector

Deploy to Manufact Cloud via the mcp-use CLI (GitHub-backed):

pnpm deploy       # runs: mcp-use deploy --org <your-org>

Or self-host anywhere with Node 20.19+ / 22.12+:

HOST=0.0.0.0 ALLOWED_ORIGINS=https://app.example.com pnpm start

index.ts                 server startup, cors, health, shutdown
src/
  config/                env parsing, capability detection, lazy proxy config
  clients/               provider API clients (serper, reddit, scrapedo)
  prompts/               optional MCP prompts for deep-research and reddit-sentiment
  tools/
    registry.ts          registerAllTools() — wires 3 tools + 2 prompts
    start-research.ts    goal-tailored brief + static playbook
    search.ts            web-search handler (with CTR-weighted URL aggregation + LLM classification)
    scrape.ts            scrape-links handler (reddit + web branches in parallel)
    mcp-helpers.ts       response builders (markdown + structured MCP output)
    utils.ts             shared formatters
  services/
    llm-processor.ts     AI extraction, classification, brief generation — primary + fallback model, always low reasoning
    markdown-cleaner.ts  HTML/markdown cleanup
  schemas/               zod v4 input validation per tool
  utils/
    sanitize.ts          strips URL/control-char injection from follow-up suggestions
    errors.ts            structured error codes (retryable classification)
    concurrency.ts       pMap/pMapSettled — thin wrappers over p-map@7
    retry.ts             exponential backoff with jitter
    url-aggregator.ts    CTR-weighted URL ranking for search consensus
    response.ts          formatSuccess/formatError/formatBatchHeader
    logger.ts            mcpLog() — stderr-only (MCP-safe)

Key patterns: capability detection at startup, description-led tool routing (no bootstrap gate), always-on structured MCP tool output, tiered classified output in web-search, parallel reddit + web branches in scrape-links, bounded concurrency via p-map, CTR-based URL ranking, tools never throw (always return toolFailure), and structured errors with retry classification.

MIT