Ledgerline

Finance-native analysis and interactive financial visualization inside ChatGPT.

Ledgerline is not a standalone finance dashboard. It is a ChatGPT plugin built as a remote MCP server with an MCP Apps UI resource.

Its role in the finance ecosystem is to provide the missing analysis layer between financial data sources and the conversation:

primary-source SEC EDGAR / XBRL research
FRED macroeconomic data
dilution analysis
filing-text search
value provenance
reusable finance calculations
interactive charts, tables, and source inspection rendered directly inside ChatGPT
a generic renderer for structured financial data already present in a conversation from another connected service or user-provided source

Ledgerline does not directly call or depend on other plugins. The model can use Ledgerline's render_financial_view tool after data from another connected source is already available in the conversation.

Tools

Tool	Purpose	UI
`lookup_company`	Resolve company/ticker to SEC identity	No
`company_financials`	Retrieve and chart SEC/XBRL metric history	Yes
`macro_chart`	Retrieve and chart FRED series	Yes
`dilution_view`	Analyze share count, SBC, and filing-language dilution signals	Yes
`search_filings`	Search recent 10-K, 10-Q, and 8-K text	No
`trace_value`	Trace a reported metric to source metadata	No
`render_financial_view`	Visualize structured finance data already in the conversation	Yes
`calculate_financial_metrics`	Period growth, index, margin, variance, CAGR	No

In-chat finance canvas

The same MCP Apps component is shared by the visual tools. It supports:

line charts
bar charts
waterfall-style views
multi-series comparisons
index-to-100 toggle
latest-value cards
tabular inspection
clickable data points
source links and notes
fullscreen mode when supported by the host

The UI follows the MCP Apps bridge rather than relying on a separate web product.

Local test

From this directory:

npm install
LEDGERLINE_USER_AGENT="Ledgerline/0.1 your-email@example.com" npm start

Then open MCP Inspector:

npx @modelcontextprotocol/inspector@latest

Connect using Streamable HTTP:

http://localhost:8787/mcp

Deploy to Render

The repository includes render.yaml.

Merge this branch.
In Render, create a Blueprint from https://github.com/Epple3k/portfolio.
Set LEDGERLINE_USER_AGENT to a descriptive SEC-compatible user agent with your contact email.
Leave OPENAI_APPS_CHALLENGE blank until the OpenAI submission portal gives you a domain-verification token.
Deploy.
Verify:
- https://ledgerline-finance.onrender.com/
- https://ledgerline-finance.onrender.com/mcp

If Render assigns a different hostname, update mcp.json before packaging the marketplace ZIP.

Connect to ChatGPT for development

After deployment:

ChatGPT → Settings → Security and login → Developer mode.
Open ChatGPT Plugins and select +.
Add the public HTTPS endpoint ending in /mcp.
Create the connection.
Start a new chat and select Ledgerline.

Useful test prompts:

“Compare NVDA revenue and operating income over the last 8 reported periods.”
“Show me the dilution picture for PLTR.”
“Chart the 10-year Treasury rate and unemployment rate together.”
“Search SOFI's recent filings for credit losses.”
“Render this as a Ledgerline chart: Revenue Jan 10, Feb 12, Mar 15; Costs Jan 7, Feb 8, Mar 9.”

Public directory / marketplace

The plugin.json file includes listing metadata and the required five positive / three negative review cases.

Before public submission you still need to:

confirm the deployed MCP hostname and update mcp.json if necessary;
merge the legal/support docs so the HTTPS GitHub URLs in plugin.json resolve;
complete OpenAI individual or business publisher verification;
record a reviewer-accessible walkthrough video and add its URL in the submission portal (or manifest);
upload a ZIP whose root is this ledgerline-plugin/ folder;
connect the MCP endpoint in the plugin submission portal;
when the portal displays its domain-verification token, set that exact value as the Render environment variable OPENAI_APPS_CHALLENGE, redeploy, and verify that /.well-known/openai-apps-challenge returns only that token as plain text;
run the automated checks and submit for review;
after approval, select Publish plugin.

OpenAI review is required before Ledgerline can appear in the universal ChatGPT/Codex plugin directory. Approval cannot be bypassed or pre-approved from the repository.

Data and safety

Ledgerline v0.1 is read-only. It does not place trades or modify financial accounts.

SEC issuers do not all use identical XBRL concepts, so the metric layer includes common concept fallbacks but should still be treated as a prototype rather than a production market-data terminal. Material conclusions should be checked against the linked primary source.

Portfolio blog email subscriptions

The portfolio Blog includes an email signup form. Subscribers are stored in a Resend Audience/Segment and the existing private publisher sends a Broadcast automatically after a new public/notes.json post is committed.

Configure these environment variables on the Render service:

RESEND_API_KEY — Resend API key with contact + broadcast access.
RESEND_AUDIENCE_ID — the Resend Audience/Segment ID used for blog subscribers.
BLOG_FROM_EMAIL — sender identity, for example Emit Rice <blog@emitrice.com>.
BLOG_PUBLIC_URL — link used in notification emails. Defaults to https://emitrice.com/?blog=1.

Before using blog@emitrice.com, verify emitrice.com in Resend and add the DNS records Resend provides.

Flow:

Visitor enters an email in the Blog header.
POST /api/blog/subscribe adds the address to the Resend list.
Publishing through /blog-admin still commits the post to GitHub first.
After that commit succeeds, the server creates and sends a Resend Broadcast.
The email contains Resend's unsubscribe URL, so opt-outs are handled by Resend.

If notification delivery fails, the blog post remains published and the publisher response includes a notification.error value for debugging.