Overview
OpenBB Agent Research OS turns market research into a repeatable, auditable agent workflow. It combines a local OpenBB backend, compact command-line tools, persistent research memory, reusable skills, and human-approved state changes so long-running agents can revisit a thesis without losing its reasoning trail.
Key Features
- Candidate radar, scorecards, catalysts, risk flags, and action inboxes
- Evidence levels that separate confirmed, inferred, speculative, and unverified claims
- Proposal-based updates with explicit approval before durable state changes
- Persistent candidate state, daily research logs, and decision history
- Compact agent CLIs for token-efficient automation and OpenBB Workspace for visual review
- Hard boundaries around trading, publishing, messaging, and other external actions
Technical Approach
The Python backend connects OpenBB-oriented data tools with the Serenity CLI, Workspace helpers, local JSON state, reusable operating skills, and scheduled heartbeat workflows. Agents can monitor markets, verify claims against quality sources, and propose evidence or thesis updates, while irreversible and money-moving actions remain behind an explicit human gate. The packaged backend includes a Pytest verification suite and example research state.