How the judge works.

From your website to your first brief

1. Confirm your email. The setup link lets you confirm the address where your briefs will arrive.
2. Check your company profile. We draft it from your public website. Review the services, countries, reference projects and constraints before approving it.
3. Start the first brief. We assess a limited selection of promising notices. Processing time and the number of matching notices vary.

How notices are assessed

1. Read the feeds. We fetch notices through four official APIs within our feed and category limits. Duplicate notices are filtered before assessment.
2. Apply the first filters. Rules check countries, categories and timing before the model runs. A missing country in your profile can exclude relevant work.
3. Assess the fit. For a selection of promising notices, a language model accessed through OpenRouter considers the available notice fields and your checked profile. It returns BID, STRETCH or SKIP, with structured reasons, risks and an effort estimate.
4. Prepare the brief. You can inspect verdicts in the dashboard. Emails have size limits, so use the dashboard when a brief cannot list every result.

What the three verdicts mean

BID: the available notice suggests a fit with your capabilities. Check the tender documents and eligibility conditions before committing.

STRETCH: a possible fit with a named gap or condition to investigate, such as a credential, reference or partner requirement.

SKIP: the notice appears outside your profile or constraints, with a reason you can challenge.

Country, minimum-value and off-limits rules can rule out work. A verdict is an assessment of the available evidence, not a guarantee that all requirements are met.

How we check quality

We use labelled development cases to test the model and rubric. These cases inform the implementation, so a score on them is not an independent measure of real-world accuracy. We do not claim a validated production accuracy percentage.

You can agree or disagree with a verdict and record why. That feedback helps us investigate errors; it does not automatically retrain a model or edit your capability profile. You can correct the profile yourself.

Known limits

• Notice coverage: category filters, date windows, missing source data and feed failures can leave gaps. We do not promise every opportunity.
• Incomplete descriptions: some published or previously stored notices contain limited text. Check the buyer’s original notice when the reasoning has too little evidence.
• Tender documents: we do not review a full pack behind a portal login. Insurance, qualifications and delivery conditions may appear only there.
• Language and model errors: nuances can be missed, and a confident explanation can still be wrong. Check material facts against the source.
• Your profile: outdated capabilities, countries or constraints affect the judgement. Review them as your firm changes.

Delivery and your data

The free offer is an initial brief plus four Monday briefs. Scheduled runs start at 7am UK time (8am in Paris); delivery follows processing. Pro is planned for weekday briefs and is not open for purchase yet. See pricing and availability.

We use your public website to draft the profile and store the account, profile, brief and feedback information needed to run the service. See the privacy policy for processing and deletion details.

Inspect and challenge a judgement

Use the dashboard to read the reasons and give feedback. If you connect an AI assistant over MCP, you can also ask about stored verdicts and request profile changes. Connection details are available in your account.