Skip to content

ginlix-ai/alternative-data

v1.0.0Apache-2.0

Signals from outside the market tape. X post search back to 2006 with profiles and reply threads, and web scraping with anti-bot bypass that returns markdown rather than raw HTML.

GitHubOpen Website

alternative-data

Two servers that look outside the market tape: X post search and general web scraping.

x_api ships with an empty env on purpose

The bearer token reaches the tool as a per-call argument read from the workspace vault (from vault import get; get("X_BEARER_TOKEN")), never from the MCP subprocess environment. In a multi-tenant deployment every workspace gets its own token that way; putting it in env would hand one host-wide token to all of them.

A single-tenant self-host that wants the convenience can add "X_BEARER_TOKEN": "${X_BEARER_TOKEN}" to the server's env in mcp.json. The tool falls back to the process environment when the argument is omitted.

scrape is ours rather than scrapling mcp

It imports scrapling as a library and runs under either mcp SDK era. The package's own scrapling mcp entrypoint is pinned to mcp 1.x and cannot start from a 2.x environment, which is the failure this server exists to avoid.