From your candidates
to a reviewable batch.
Submit the headlines your agent read and the contracts it selected. We check the relationship and return a structured batch you can use in your own evaluation.
1. Arrange your sample
Request a sample review with your team’s product URL, what the agent does, and the selection problem you’re seeing. We confirm eligibility and one of the three free sample slots before you submit material.
2. Supply headlines and candidate markets
Use the template for up to five inputs in a free sample or 25 in a paid pilot. Each input accepts up to three candidates. Include the original source URL and publication time when known, plus venue and market ID or URL. Start with public US policy and macroeconomic sources.
Do not send credentials, private account information, or material you cannot share for AI-assisted review. Delivery and payment are arranged directly, outside this website.
3. Read the reviewed output
| Field | Meaning |
|---|---|
schema_version | Output format version, currently 1.0.0. |
batch_id, reviewed_at | Batch identity and review timestamp in UTC. |
retrospective | Whether this is a retrospective review. Historical examples are not point-in-time backtests. |
items[].input | Text, source URL, and known publication timestamp. Unknown timestamps are null. |
candidates[] | Venue identity, question, raw API end date, rules source, verification timestamp, verdict, reason, references, and missing information. |
"decision": "uncertain", "reason": "This source requests a feasibility study; it does not establish creation of the agency.", "missing_information": [ "An executive action or law establishing the agency within the contract window." ]
Decision fragment only. Download the complete example output to see a valid batch.
What the decisions mean
- Relevant: the input directly addresses the named contract’s condition. Evidence against a Yes outcome can still be relevant.
- Irrelevant: the supplied input addresses a different event, authority, metric, or time window. Indirect forecasting value is a separate question.
- Uncertain: missing or ambiguous evidence prevents a supported decision. A proposal alone does not establish implementation.
Dates and unavailable evidence
The deadline field preserves the venue API’s raw endDate. It is displayed as “venue end date (metadata).” It can differ from the actual event or rule cutoff. The explanation and rules references identify that distinction; never use this field alone to decide settlement.
Unknown information stays null. Missing contract rules or inaccessible evidence produce an uncertain review with explicit missing information. A deliberately incomplete demonstration is labeled as such; it is not a claim that the real market is unavailable.
Delivery and limits
The $200 pilot covers 25 inputs with up to three candidate contracts each, delivered within two business days after scope is agreed and complete inputs are received. Reviews are prepared with AI assistance and checked before delivery. We return JSON directly and keep customer batches outside this public website.
This release provides a batch format and arranged review service. There is no production API, MCP integration, continuous feed, trading execution, or claimed trading edge.
Validate and integrate
Use the downloaded JSON Schema (Draft 2020-12) in your existing validation tools. Preserve identifiers so you can compare reviews with your agent’s output. Structural validation cannot verify the truth of a source or replace contract review. The public examples and format may be used in your internal evaluation; original publications and venue content retain their respective rights.