Record claims and falsifiers
The investor states why the business deserves attention and what would prove each reason wrong.
Open-source engine · Python library
Suvi’s engine for accountable thesis monitoring.
AgenticThesis tests an investor-written thesis against new company filings—with exact citations, Human Review, and version history.
What the engine does
New evidence can challenge a thesis without silently rewriting it.
The library connects each disclosure, citation, proposed change, review decision, and accepted thesis version.
The engine proposes. The investor decides what becomes thesis history.Open Getting Started
Disclosure-to-review workflowAvailable in the open-source alpha
Thesis
Preserve official evidence
Selected SEC and configured issuer IR disclosures retain their original artifacts and provenance.
Propose a thesis delta
Relevant evidence supports, weakens, possibly invalidates, or leaves a claim unknown.
Require Human Review
Only an approved change can create the next authoritative thesis version.
Available today
One open-source engine, inspectable at every boundary.
Python engine
AgenticThesisEngine embeds the workflow in another Python application.
Self-hosted service
agentic-thesis serve provides the local UI, HTTP API, and monitoring loop.
Exact citations
Accepted quotes resolve to preserved artifacts, pages, or exact character ranges.
Checkpoint and replay
Resume, replay, and conflict protection preserve long-running work and version ownership.
Engine boundary
Professional thesis monitoring—not a complete investment service.
AgenticThesis deliberately leaves valuation, portfolio construction, and investment responsibility with the investor.
- 01No price predictionOut of scope
It does not forecast stock prices or market direction.
- 02No Buy, Sell, or Hold instructionOut of scope
It tests reasons; it does not own the investment decision.
- 03No automated valuation or position sizingOut of scope
Those judgments remain outside this engine.
Documentation · source · distribution