I built a self-hosted open-source MCP server that gives any local LLM real financial data — SEC filings, 13F, insider & congressional trades, short data, FRED

Posted by DanielAPO@reddit | LocalLLaMA | View on Reddit | 11 comments

One thing missing when running local models as agents: real, current data. So I built Equibles — a self-hosted MCP server that scrapes and serves public U.S. financial data and exposes it as MCP tools, so any MCP-capable client (Claude Code/Desktop, Cursor, or your own local-model agent loop) can query it directly.

No cloud dependency, no API keys, no telemetry — it all runs on your machine.

What it serves:

SEC filings (10-K/10-Q/8-K) with full-text search
13F institutional holdings, insider (Form 3/4) and congressional trades
FINRA short volume / short interest, SEC fails-to-deliver
FRED economic indicators, CFTC futures positioning, CBOE VIX/put-call
Daily prices + technical indicators

I'm the developer. Feedback and feature suggestions are very welcome.

Repo: https://github.com/daniel3303/Equibles (leave a star if you liked it :) )

[-]

CaptainProud4703@reddit

this is pretty good,

rm-rf-rm@reddit

Nice! does it download stuff on demand or pre-cache a bunch of things?

And given this is completely local, what is the benefit of having the interface betweent the agent and this an MCP, instead of a CLI or direct API calls?

Training-Web7861@reddit

wait so local models can actually do financial analysis now?

Yorn2@reddit

I see you are getting your data from Yahoo Finance. Have you ever had them block you off or data-limit you for hitting their API/services?

DanielAPO@reddit (OP)

Hi, never happened, the system has a rate limiter

ByteDinosaurs@reddit

this fills a real gap tbh

local models as financial agents has always been hobbled by stale data. congressional trades and 13F holdings via MCP with no api keys or telemetry is a legitimately useful combo

starring this

a_beautiful_rhind@reddit

Time to get into investing and lose some money!

jake_that_dude@reddit

this is the right shape for local agents. the thing i'd add early is a provenance layer: every answer should carry accession_number, filing date, source URL, and retrieval timestamp. financial data gets stale/weird fast, and without that audit trail the LLM can sound confident while mixing a 10-Q, a 13F, and yesterday's price feed into one fake narrative.