Agents Compare Graph Research Tools Method About Login Get Pro →

Claude alone vs Claude + ForcedAlpha

One describes a category. The other traces the dependency. Same question. One answer is confident. The other is checkable. A live, side-by-side demo: pick a question (or ask your own) and the same underlying model answers it twice — once with no graph context, once grounded in the ForcedAlpha dependency graph.

Pick a question, or ask your own — both sides run live

There is no pre-recorded example on this page. Click one of the sample questions — including "impact on a book" seeded with a real ticker list (LMT, NOC, RTX, MP, TSM, NVDA, UAMY) against a rare-earth export-control scenario — or type your own question about any supply chain, material, or company. Both sides run the same live request: Claude alone, and Claude grounded in the ForcedAlpha dependency graph.

Claude alone tends to answer in generic, directional prose — no named companies, no traced path, and (when a portfolio is involved) an upfront disclaimer that it has no real-time market or position data. Claude + ForcedAlpha plans, queries, and traverses the actual dependency graph before answering, naming the specific companies and materials involved, tracing the hop-by-hop path, and stating plainly when a ticker has no path in the graph rather than guessing.

How the live answer is shown

The graph side can take longer than the raw side because it is actually querying and traversing the graph before it generates — that lag is itself part of what you're seeing. Its answer renders core-insight first, with the full hop-by-hop trace collapsed behind a "show the trace" toggle and citations last.

General AI describes the world. ForcedAlpha traces what the world depends on.

Want your own agent to have this capability directly? Give it the graph via MCP → forcedalpha.com/agents. Or run the same comparison yourself at forcedalpha.com/demo/live.

Public-source graph. Every edge carries its evidence grade. Refuses when the graph cannot support the claim.

Claude alone vs Claude + ForcedAlpha

One describes a category. The other traces the dependency.

Same question. One answer is confident. The other is checkable. Click a question below, or ask your own.

These are questions a terminal can't answer — they require traversing the graph.

Pick a question — or ask your own. Both run live.

Click a question above, or type your own, to run both sides live.

General AI describes the world. ForcedAlpha traces what the world depends on.

Want your own agent to have this capability directly, on any question — not just this demo? → Give it the graph.

Public-source graph. Every edge carries its evidence grade. Refuses when the graph cannot support the claim.