Hello Everyone — I've posted here a few times over the past year to share the site I've been building ([www.dfin.pro](http://www.dfin.pro)). I'm a fundamentals-based value investor, and I'm developing this site exactly for people like me. There are a lot of trading-focused sites, but that's not what this is.
Thanks to the admins for keeping my posts live here so far. I hope it's okay that I'm back to share some more updates: I've opened up direct access to our database via MCP and an API.
Quick background: I started with a chatbot on the site, then added quarterly research reports as many people said that they didn't have a single quality place to get reliable research/coverage on the names they're interested in. We've published nearly 1,600 reports so far on 105 companies.
Thank you to everyone for the valuable feedback you have provided so far.
Now seeing how usage for AI has shifted over the last several months shifted to subscription plans, I felt opening up an the site through an MCP/API is the logical next step. It's the same data, but headless, so you can pull it straight into whatever agent or workflow you're already using.
I'd love feedback from people here. In exchange, I'm happy to give free access to full functionality for an extended period. Please just comment or DM me.
Here's what you can do on the site right now:
* Screening - lots of advanced filters plus some visualization tools that make it easy to spot outliers.
* Financials - \~15 years of standardized and as-reported financials for US stocks. I do the standardization myself, and every data point links back to the actual filing. You can download to Excel as well.
* Research reports - quarterly reports generated after earnings.
* Notebook - Your personal research repository. Works best through the MCP.
The MCP gives you all of that in a headless way, plus a few things that are genuinely hard to do otherwise:
* Semantic and keyword search across filings (\~3,800 US companies and growing), transcripts, and our reports. A lot of information lives in slides and charts and normally just gets lost. The search functionality on the site doesn't miss that information.
* Full natural-language screening, not a stripped down version of what's on the site.
* Semantic screening. Narrow down a list of stocks with quant filters, then search for phrases like "strong growth," "new product segment," or "decreasing competition" and find the companies actually saying that.
* Financials pulled directly with a link back to the source (metrics coming next).
* Macro data and stats from public sources.
* Maintaining your notes as an actual living knowledge base. Each note you add to the site becomes a part of an interlinked corpus. With the power of a graph database, searching for interconnected threads becomes easier. The notes are not "atomic" - every note is linked to other relevavnt notes, companies, filings (Down to the exact snippet), and transcripts. New input is scanned against the existing corpus to see whether it strengthens, challenges, or fills in gaps. Never lose any data point, or thought and insight that you've generated. Using an agent for drafting my notes has meant it's taken away the laziness factor in doing this.
I also spent a lot of time trying to make the MCP token efficient. Data comes into context in small snippets instead of dumping whole documents, and financials are pre-extracted so the agent isn't hunting for numbers. And since the quarterly reports already exist, your agent can build on that basic research instead of redoing it every time (how the quarter looked, whether guidance moved, balance sheet health, key questions from the call, etc.) rather than burning tokens on the repetitive stuff.
The API is live too. I'm still tightening up the surface so it needs a little more work to stabilize the parameters. It works though if you want to try it!
I hope that sheds some light on what I've been building. Please drop a comment or DM if you want access, happy to set you up.
Thank you very much!