@moneyline/mcp-server
v0.1.1
Published
Moneyline MCP server — AI agent tools for financial document intelligence
Maintainers
Readme
@moneyline/mcp-server
Model Context Protocol server for Moneyline — exposes the API as tools and resources for AI agents (Claude, and any MCP-compatible client).
Tool/resource paths are verified against the OpenAPI spec by bun run sdk:parity
(from the repo root), so they can't silently drift from the API. Run
tools/list against the server for the live, authoritative tool catalog.
Install
npm install -g @moneyline/mcp-server # provides the `moneyline-mcp` binaryOr run it without installing, using npx @moneyline/mcp-server.
Transports
Selected by MCP_TRANSPORT:
| Mode | Use | Auth |
| ------- | ------------------------------------- | ------------------------------------------------- |
| stdio | local (Claude Desktop, CLIs); default | MONEYLINE_API_KEY env — one key per process |
| http | deployable hosted server | per-request Authorization: Bearer <key> header |
stdio (local)
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"moneyline": {
"command": "moneyline-mcp",
"env": {
"MONEYLINE_API_URL": "http://localhost:3000",
"MONEYLINE_API_KEY": "ml_live_xxx"
}
}
}
}http (hosted)
MCP_TRANSPORT=http MCP_PORT=3333 MONEYLINE_API_URL=https://api.moneyline.co \
bun run startRuns statelessly: each POST /mcp is handled with a fresh server bound to the
caller's API key from the Authorization: Bearer <ml_live_*> header, so one
deployment serves many tenants. GET /health is an unauthenticated liveness
probe; requests without a key get 401.
Develop
bun run dev # stdio, tsx watch
bun run test # vitest
bun run build # tsup -> dist/See .env.example for all variables.
License
MIT
