How to Pull SEC 8-K Filings and Earnings Call Transcripts as JSON

I spent a weekend once writing a parser for SEC 8-K filings. It worked until a company arranged its exhibits some new way, and then it silently mislabeled the event. Earnings call transcripts were worse: there is no official source to parse, just third-party pages with their own layouts. Keeping both alive across thousands of US filers is not a weekend project, it is a job.

The Earnings Call Transcript API is a hosted scraper that returns parsed 8-K filings and speaker-tagged earnings call transcripts as structured JSON, so you skip the parsing entirely.

What it returns

Two record types. For filings: item codes and item names, the extracted press release, isolated guidance sentences, a sentiment score, and the source EDGAR url. For transcripts: participants with roles, prepared remarks, and analyst question-and-answer pairs matched to the executives who answered. Every record links back to its filing, so a claim is always checkable. There is no audio, and no 10-K or 10-Q fundamentals; this is an events-and-calls feed.

Run it in three steps

Open the Actor, click Try for free, enter one or more tickers, and pick a dataType: filings, transcripts, or both. Run it, then export the records as JSON, CSV, or Excel. That is the whole loop.

Use it as a REST API

Every run is callable over HTTP, so a scheduled job can pull fresh records without a browser:

POST /v2/acts/johnvc~earnings-call-transcript-api/runs
{ "tickers": ["AAPL"], "dataType": "both", "filingsLimit": 5 }

The full parameters are in the Apify API docs. Output comes back as one record per filing or transcript.

Wire it into Claude with MCP

Through the Model Context Protocol, Claude, Claude Code, and Cursor can call the Actor as a tool mid-conversation. Ask "summarize Apple's latest 8-K and pull its guidance language" and the answer runs against the real filing instead of the model's memory, with no scraping code to write. You can read more about Claude at claude.ai.

What people build with it

The examples tab has runnable tasks for each: red-flag screeners that full-text search every US filer for material weakness and going concern language, event feeds off 8-K item codes, a pre-matched Q&A dataset for LLM and RAG work, and ticker transcripts for AAPL, NVDA, and TSLA. Event data pairs well with the Google Finance API for prices and the Congress Financial Disclosures API for another primary-source signal.

FAQ About Scraping SEC Filings and Earnings Calls

EDGAR is free, so what does this scraper actually add?

EDGAR returns documents; this scraper returns data. You pay for the parsing: item codes tagged, press releases extracted, guidance sentences isolated, Q&A matched to speakers, plus transcripts that EDGAR never had.

How do I call the API and what does a record cost?

POST to the Actor's runs endpoint with your tickers and dataType, then read the dataset as JSON, CSV, or Excel. Billing is one event per record returned with no start fee, so a hundred records cost about a cent and a quarterly refresh of a 50-ticker watchlist runs about the same.

Can Claude or another AI agent run this through MCP?

Yes. Connect the Apify MCP server and the Actor becomes a callable tool, which makes filings and transcripts available to an agent as grounded context instead of recall, no code required.

Can an AI agent pay for this scraper in USDC with x402?

Yes. The Earnings Call Transcript API supports agentic payments via the x402 protocol, so AI agents and MCP clients can pay for runs in USDC on Base with no Apify account or API token. Point your agent at the Apify MCP server and it can discover, pay for, and run the scraper autonomously; the Apify x402 announcement has the details.

Where does the scraper's coverage end?

Transcripts cover roughly 1,000 to 1,500 mostly large and mid cap US companies per quarter, with an archive back to about 2007, so a thin micro cap may have filings but no call. Sentiment is a deterministic finance-dictionary score, not an LLM read.

Issues get a same-day reply. Point the Earnings Call Transcript API at your watchlist and get 8-Ks and calls back as records you can compute on.