Description
< #Multi-Agent #Investment Analysis #Stock Analysis #AI Agent #Value Investing #MCP #Open Source Software
Augur is an open-source multi-agent investment analysis system built on a simple premise: let more than a dozen virtual investing masters reach their verdicts independently, then merge them through a weighted consensus mechanism. Rather than asking one model for a single answer, it deliberately preserves disagreement — value, growth and macro perspectives each argue their own case, and what you see in the end is a full picture with both consensus and dissent.
On the data side it pulls real filings from SEC EDGAR, and the front end is a Bloomberg-style dark dashboard where analysis, committee voting and bull-versus-bear debate all live in one place. It suits anyone studying how AI agents collaborate on decisions, or looking to add an automated cross-check layer to their own research workflow.
Independent multi-agent analysis:18 virtual investing masters each hold a distinct philosophy and evaluation framework, scoring the same security independently so no single viewpoint introduces systematic bias.
Weighted consensus:Agent conclusions are not simply averaged — they are combined by weight into a final verdict while the points of disagreement stay visible for human review.
Real filing data:Connects to SEC EDGAR for genuine disclosure data, grounding the analysis in public filings rather than model guesswork.
Bloomberg-style dashboard:A professional dark-themed web interface with separate modules for stock analysis, the investment committee, bull-bear debate, history and side-by-side comparison.
Bull-bear debate view:Lays the bullish and bearish arguments side by side so that the disagreement itself becomes readable information instead of being flattened into an average.
MCP tool integration:Ships 13 MCP tools, so the full analysis chain can be invoked directly from Claude Desktop and other MCP-capable clients.
Full CLI:Beyond the web interface there is a complete command-line entry point, convenient for scripted invocation and batch analysis.
Installable PWA:The web app can be installed to the desktop as a progressive web app for a near-native experience.
Open source:Released under the MIT license on GitHub, so developers can inspect the implementation, contribute, or adapt it to their own needs.
Augur is an open-source multi-agent investment analysis system built on a simple premise: let more than a dozen virtual investing masters reach their verdicts independently, then merge them through a weighted consensus mechanism. Rather than asking one model for a single answer, it deliberately preserves disagreement — value, growth and macro perspectives each argue their own case, and what you see in the end is a full picture with both consensus and dissent.
On the data side it pulls real filings from SEC EDGAR, and the front end is a Bloomberg-style dark dashboard where analysis, committee voting and bull-versus-bear debate all live in one place. It suits anyone studying how AI agents collaborate on decisions, or looking to add an automated cross-check layer to their own research workflow.
Features
Independent multi-agent analysis:18 virtual investing masters each hold a distinct philosophy and evaluation framework, scoring the same security independently so no single viewpoint introduces systematic bias.
Weighted consensus:Agent conclusions are not simply averaged — they are combined by weight into a final verdict while the points of disagreement stay visible for human review.
Real filing data:Connects to SEC EDGAR for genuine disclosure data, grounding the analysis in public filings rather than model guesswork.
Bloomberg-style dashboard:A professional dark-themed web interface with separate modules for stock analysis, the investment committee, bull-bear debate, history and side-by-side comparison.
Bull-bear debate view:Lays the bullish and bearish arguments side by side so that the disagreement itself becomes readable information instead of being flattened into an average.
MCP tool integration:Ships 13 MCP tools, so the full analysis chain can be invoked directly from Claude Desktop and other MCP-capable clients.
Full CLI:Beyond the web interface there is a complete command-line entry point, convenient for scripted invocation and batch analysis.
Installable PWA:The web app can be installed to the desktop as a progressive web app for a near-native experience.
Open source:Released under the MIT license on GitHub, so developers can inspect the implementation, contribute, or adapt it to their own needs.
