AIニケちゃん Knowledge Search MCP Server

This is an unofficial tool. This project is not approved, affiliated with, or sponsored by AIニケちゃん, its operators, X, OpenAI, or any related companies or organizations. Rights to names, posts, websites, and other sources belong to their respective rights holders.

This is a Next.js / Vercel server for searching public information about AIニケちゃん from MCP (Model Context Protocol) clients. Queries are vectorized by OpenAI Embeddings via the Vercel AI Gateway, and search results include source URLs and post timestamps.

What Not to Include in the Public Repository

The following will never be committed to this repository:

• Collected X post data and derived text

• Search metadata including post content and generated vectors

• .env-style environment variable files, API keys, and authentication tokens

These are excluded via .gitignore and .vercelignore. In public deployments, the generated index is loaded from Private Vercel Blob, so post data is not included in the repository or Function artifacts.

Public Demo and Security

This project exposes the MCP endpoint without authentication for a short-term public demo. In Vercel, register AI_GATEWAY_API_KEY and the storage settings added when connecting Private Blob. Do not store the values in GitHub, client code, or logs.

AI_GATEWAY_API_KEY=
BLOB_READ_WRITE_TOKEN=

Since third parties can also run searches, AI Gateway usage fees will be incurred. Keep the public period short, configure Vercel's WAF and rate limits, and monitor usage. Stop the deployment after the demo ends.

Values that look like credentials have been detected in the .env file in this working folder. Before adding them to Git, confirm they are not currently in use, and rotate the keys if necessary.

Local Development

npm install
cp .env.example .env.local
# .env.local に OPENAI_API_KEY を設定
npm run build:index
npm run dev

The local search endpoint is http://localhost:3000/api/mcp. Generating the index requires the excluded X post data locally.

Deploying the Index to Vercel Blob

1. From your Vercel project's Storage, create a Blob store with Private access and connect it to the project.

2. Fetch the Vercel environment variables locally and set BLOB_READ_WRITE_TOKEN in your local .env.local.

3. After generating the index locally, run the following:

npm run upload:index

indexes/metadata.json and indexes/embeddings.f32 are uploaded to Private Blob. The Vercel environment loads these two files, while the local environment loads the files in src/data as before. Only if you change the paths inside Blob, set VECTOR_INDEX_METADATA_PATH and VECTOR_INDEX_VECTORS_PATH to the same values in both local and Vercel environments.

MCP Tools

• search_nikechan_knowledge: Semantic search across X posts and the official website

• search_x_posts: Semantic search over X posts only

• search_hybrid: Search combining keyword matching and semantic search

• search_keywords: Keyword search

• search_timeline: Search by period or year/month

• get_popular_posts: Search sorted by reaction count

• get_x_post / get_x_thread: Retrieve a post or thread

• x_posts_dataset_info: Check the record count and structure of the index

The results are reference information based on public information. Accuracy, completeness, and freshness are not guaranteed, so please verify the original sources at the returned URLs in your answers.

Verification

npm run lint
npm test
npm run build

Known Pre-Deployment Checks

The dependency audit still shows high-severity vulnerabilities in the Next.js family. Fixing them involves breaking changes when upgrading to Next.js 16, so this repository does not auto-update. Before production deployment on Vercel, update Next.js after confirming compatibility, and run npm audit --omit=dev --audit-level=high again.