Multi-agent forecasting

See the market from seven directions.

Forecast AI brings Polymarket, Kalshi, live research and seven specialized agents into one explainable probability.

Explore the product
Forecast AI workspace
Selected market

Will the Fed cut rates before year-end?

Polymarket 42%Kalshi 44%
Independent agent run7 / 7 complete
Market
49%
News
52%
Macro
55%
Social
58%
Reddit
61%
Research
64%
Risk
67%
Consensus
58%

Evidence-weighted probability after measuring agreement and contradictions across seven agents.

Market
42%
Observed edge
+16%
Seven agents, clearly separated

Independent reasoning before consensus.

Market, news, macro, social, Reddit, research and risk analyze the question from different angles. Only after all seven finish does the consensus engine combine their evidence and disagreement.

Market agent
Price, liquidity and venue signals
News agent
Current reporting and primary evidence
Macro agent
Economic and policy context
Social agent
Narrative velocity and sentiment
Reddit agent
Community-specific discussion
Research agent
Deep evidence synthesis
Risk agent
Contradictions and uncertainty
Consensus engineAfter 7 analyses

Combines seven independent outputs only after their analysis is complete.

Evidence, not a black box

Know where every signal came from.

Citations remain connected to the agent and research provider that supplied them. Enabled integrations can include Tavily, FactsAI and Falcon / Polygun.

Provider status is shown honestly in the workspace.
Evidence provenance
Research layers

Provider information stays attached to evidence instead of disappearing inside one generated answer.

Tavily
Live web research
Available
FactsAI
Partner research connector
Connector ready
Falcon / Polygun
Market intelligence layer
Token required
Source-level attribution

Each result can expose its citation, provider and agent context. Availability follows the integrations enabled in the environment.

Forecast track record
Proof-ready
Forecast payload hash
0x8b7f2a...59d4c1

Deterministic hashes are generated now. Robinhood Chain publication remains a separate deployment step.

Offchain scoring active
0.14
Brier score
lower is better
AccurateUncalibrated
From forecast to track record

Accuracy that can be measured.

Forecast payload hashes and offchain Brier scoring are ready today. Publishing those proofs to Robinhood Chain is the next deployment phase, not a feature we pretend is already live.

Available now
Hash + Brier history
Next phase
Onchain registry
Forecast AI workspace

One workspace for markets, agents and evidence.