Open source
Open source finance is here and it cannot be stopped. A vendor can decline to build the connector. It cannot stop a doctoral student from publishing one.
25open source projects for finance research
14of them updated in the month to 10 Oct 2026
544kGitHub stars between them
Eight public projects connect WRDS to AI agents. Every one was created in 2026.
Build the dataset yourselfPublic filings, statistics and call audio are a few open tools away from a research panel. How to do it
Start from someone's skillFactor construction and reproducibility checks already exist as files you can drop into a repository. How skills work
Run the whole loopIdea, code, backtest and write-up, from projects with tens of thousands of stars. The pipelines guide
7 of 10 things you query a platform for have a free source
As of 9 October 2026| You need | Usually licensed as | Free at the source | Our call |
|---|---|---|---|
| Accounting fundamentals | Compustat | SEC Financial Statement Data Sets and the XBRL company facts APIUS filers in the XBRL years. No standardized items and no earlier history. | Free for part of it |
| Filing text and metadata | A vendor's filings product | EDGAR APIs, read with edgartools (MIT license)10-K, 10-Q, 8-K, insider forms and 13F in one Python library. | Use the free one |
| Institutional holdings | Vendor 13F files | SEC Form 13F Data SetsThe filings the vendor files are built from. | Use the free one |
| Insider trades | Vendor insider files | SEC Insider Transactions Data SetsFilings by executives, directors and large holders. | Use the free one |
| Mutual fund portfolios | Vendor fund holdings | SEC Form N-PORT Data SetsMonthly portfolio holdings as the funds report them. | Use the free one |
| Factor and anomaly returns | Factor files on a data platform | Open Source Asset Pricing and Global Factor DataSignals replicated from the literature with code, and 153 factors across 93 countries. | Use the free one |
| Earnings call text | Capital IQ transcripts | Company webcasts and open speech recognitionRecent calls for your own sample. Nobody can rebuild the history. | Free for part of it |
| US stock returns | CRSP | Nothing open matches it | Keep the license |
| Analyst forecasts | I/B/E/S | No free equivalent a search turns up | Keep the license |
| Options and intraday trades | OptionMetrics, NYSE TAQ | No free equivalent a search turns up | Keep the license |
Rebuild what is free, license what is scarce, and take the list of what you can rebuild into the renewal meeting.
WRDS, wired up by researchers
Nobody waited for an official connector.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| wrdswharton | The official Python client for WRDS. Runs SQL against WRDS and returns a dataframe. | BSD-3-Clause | 160 | |
| claude-wrds-publicpiotrek-orlowski | WRDS database skills for Claude Code, with one agent for PostgreSQL queries and one for TAQ jobs. | None listed | 30 | |
| wrds-factor-construction-skillfqybaba | Claude Code skill that builds academic factors from Compustat and CRSP in five steps. | MIT | 1 | |
| wrds-mcp-servermounist | MCP server that exposes CRSP, Compustat and IBES as tools for Claude and other models. | None listed | 1 | |
| wrds-research-mcpSharpma | A WRDS MCP server for research. | MIT | 1 | |
| wrds-mcphosseinzk | Claude plugin and MCP server that searches the WRDS catalog, checks what is available and pulls the data. Says it enforces the WRDS AI policy. | MIT | 0 |
Skills and workflows
Procedures other researchers have already written down.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| applied-micro-skillskennethkhoocy | Claude Code and Codex skills for empirical applied-micro research, including reproducibility auditing. | MIT | 22 | |
| corbis-literature-starter-kitAgentic-Assets | A Claude Code project for exploring literature and brainstorming research ideas. | MIT | 11 |
Public data for agents
Filings, statistics and markets with no license in the way.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| edgartoolsdgunning | Reads SEC EDGAR filings in Python: 10-K, 8-K, XBRL financials, insider forms and 13F holdings, without a paid filings product. | MIT | 2,781 | |
| sec-edgar-mcpstefanoamorelli | SEC EDGAR filings as tools an agent can call. | AGPL-3.0 | 368 | |
| FTShare-MCPFTShare-Lab | Chinese market data and factors for agents. | MIT | 250 | |
| Equiblesdaniel3303 | Self-hosted server for SEC filings and XBRL financials. | AGPL-3.0 | 230 | |
| fred-mcp-serverstefanoamorelli | FRED economic data as tools an agent can call. | AGPL-3.0 | 124 | |
| us-gov-open-data-mcplzinga | More than 40 US government data APIs behind one server. | MIT | 112 |
Research automation
Systems that run most of the loop.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| qlibmicrosoft | Quant research platform with data, models and backtests in one framework. | MIT | 49,244 | |
| RD-Agentmicrosoft | Agents that propose factors and models, write the code and backtest it. | MIT | 14,870 | |
| AI-ScientistSakanaAI | Takes a research idea to a written paper with no person in between. | Custom | 14,686 | |
| AgentLaboratorySamuelSchmidgall | Literature review, experiments and a report from one research idea. | MIT | 5,895 |
Replication and factor data
The benchmarks your own results get compared with.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| CrossSectionOpenSourceAP | Code and data to reproduce published cross-sectional return predictors. | GPL-2.0 | 1,064 | |
| ReplicationCrisisbkelly-lab | The code behind the global factor data of Jensen, Kelly and Pedersen. | None listed | 377 | |
| websitetidy-finance | Source for the open textbook on reproducible empirical finance. | Custom | 119 |
Models and agents
The tools underneath.| Project | What it does for a researcher | License | Stars | Updated |
|---|---|---|---|---|
| opencodeanomalyco | The open source coding agent. You choose the model provider. | MIT | 212,390 | |
| codexopenai | OpenAI's coding agent for the terminal. | Apache-2.0 | 128,390 | |
| whisperopenai | Speech recognition. Turns earnings call audio into text. | MIT | 110,247 | |
| finBERTProsusAI | Financial sentiment model. Small, old and still a common baseline. | Apache-2.0 | 2,249 |
Stars and dates are from GitHub. "None listed" means the repository has no license file, so ask the author before you reuse the code.