MCP TechTrend

What salwks/mcp-techTrend MCP server does

The salwks/mcp-techTrend MCP server collects information from academic, software, model, and medical-device sources and makes the results available as MCP tools. Supported sources include arXiv, PubMed, GitHub, Hugging Face, HF Daily Papers, FDA 510(k) clearances, and FDA medical-device recalls.

Source-specific tools cover recent arXiv papers, field-aware arXiv searches, PubMed searches with abstracts, GitHub trending pages and repository searches, Hugging Face models, datasets and spaces, HF Daily Papers, recent 510(k) clearances, and recalls with class filtering. The two aggregation tools combine enabled sources into either a bullet-list digest or a newspaper-style briefing. The briefing can accept an optional topic and is intended for requests such as a weekly news report.

How it works

The server reads its settings from the run.py launcher and registers tools according to the enabled-source configuration. Disabling a source removes that source's individual tools from the MCP tool list, while the digest and briefing tools remain available and skip disabled sources.

ArXiv categories can be assigned weights. Results use a per-category round-robin approach so high-volume categories do not completely dominate smaller categories. Search tools support recent-day filtering where applicable. Repeated and concurrent requests use a TTL cache and request coalescing to reduce duplicate upstream calls.

The salwks/mcp-techTrend MCP server performs read operations against the listed services. It does not require tokens for normal anonymous access, but optional GitHub, Hugging Face, NCBI, and openFDA tokens can increase rate limits. Token values are not returned by the configuration tool.

Setup and configuration

Installation is a local Python setup: clone the repository, create a virtual environment, and install requirements.txt. Claude Desktop is configured to run the virtual environment's Python executable with run.py as its argument. The launcher should be used instead of starting trends_mcp.py directly because it prepares the configured settings before handing off to the server.

Configuration can be edited interactively with python configure.py. The interface can select active sources, assign arXiv categories and weights, set the default PubMed query, manage optional tokens, display the current configuration, and restart stale MCP processes. --show displays settings and --restart terminates stale MCP processes. Settings can also be edited directly in the run.py settings block or changed through the trends_set_* MCP tools.

Changes take effect when the server is spawned again. After editing settings or using a configuration tool, restart Claude Desktop or restart the relevant MCP process.

Tools and capabilities

• Search arXiv by category, submission date, keywords, or field syntax such as ti:, au:, abs:, and cat:.
• Search PubMed using MeSH terms and field tags, with abstracts retrieved through efetch.
• Browse GitHub trending HTML and search repositories through the GitHub Search API.
• Inspect HF Daily Papers, Hugging Face models, datasets, and spaces.
• Retrieve recent FDA 510(k) clearances and medical-device recalls.
• Produce topic-based multi-source digests and newspaper-style briefings.
• Inspect or change enabled sources, arXiv category weights, the default PubMed query, and optional provider tokens.

Limitations and notes

The server depends on the availability and rate limits of the upstream public services. GitHub trending is obtained by scraping the trending page, while GitHub search uses the Search API. HF Daily Papers replaces the sunset Papers with Code API in the source list.

Optional credentials are rate-limit boosters rather than prerequisites. GitHub tokens should use no repository scope because the server only reads data. The configuration interface labels are in Korean, although its commands and presets are in English. The server is documented for Claude Desktop; compatibility with other MCP clients is not established by the supplied material.