An equity research agent that queries SEC filings directly via filesystem tools instead of vector databases.
Ditch the vector DB for a ReAct agent that reads 10-Ks, 10-Qs, and 8-Ks like a human analyst.
This tool is a research agent designed for finance professionals who need direct, verifiable answers from SEC filings. It replaces complex RAG pipelines with a simple filesystem-based approach, allowing the agent to browse, grep, and read documents directly.
The agent uses a ReAct loop to perform filesystem operations like ls, read_file, and grep on a local corpus of markdown-converted SEC filings. By avoiding chunking and semantic search, it ensures that every cited figure is retrieved directly from the source document.
It suits quantitative researchers and analysts who prefer deterministic retrieval over probabilistic vector search. It is not for those who want a plug-and-play SaaS, as it requires managing a persistent filesystem and periodic SEC data ingestion.
Uses standard terminal tools like grep and ls instead of vector embeddings.
Includes a background watcher that automatically fetches new filings from EDGAR.
Processes raw SEC HTML into clean markdown for reliable LLM parsing.
Integrates Tavily for real-time news to supplement historical filing data.
The tool is free to use under its license. You are responsible for all infrastructure costs, including hosting, database storage, and API usage fees for the LLM and news search.
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
These reviewed AI tools do the same job instantly — from our scored directory.
The developers found that modern LLMs are capable of navigating a filesystem and performing grep searches more accurately than traditional RAG pipelines.
It currently includes a canonical list of 300+ US tech companies, but can be expanded if you manage the ingestion pipeline.
See Zekai first in Google