How the Evaluator Works

5 steps from RFT to recommendation. No mock data, all real engine.

Extract Criteria

Check Compliance

Detect Deviations

Score Bids

Generate Report

Architecture

The Evaluator runs on a Node.js MCP server with jurisdiction-specific rulesets (Federal, NSW, VIC, QLD). Each tool in the pipeline chains via an in-memory evaluation session:

extract_criteria → check_compliance → detect_deviations → score_bids → generate_report

A Human-in-the-Loop checkpoint sits between compliance checks and scoring — flagged items go to a procurement officer for review before final scores are applied.

1. Extract Criteria

Generate the compliance checklist for a procurement.

Select jurisdiction (Federal, NSW, VIC, QLD), enter the estimated contract value, and pick the entity type (corporate/non-corporate). The engine applies the relevant Commonwealth Procurement Rules (CPRs) or state-level equivalent to generate a tailored set of evaluation criteria — from core principles like Value for Money down to compliance gates like ISO 27001, WHS, and financial viability.

2. Check Compliance

See how a bidder stacks up against each criterion.

Enter a bidder name and paste evidence for each compliance gate in `gate_id: evidence text` format. The engine cross-references each gate's evidence pattern against your input — flagging compliant, non-compliant, or needs-human-review. Mandatory gates that fail block the bidder from proceeding.

3. Detect Deviations

Spot conditional language, exceptions, and risks.

Analyses the same evidence for red flags — conditional wording (`alternatively`, `subject to`), direct exceptions (`take exception to`), and proposal variations. Each deviation gets a severity rating (high/medium/low) so evaluators know where to focus.

4. Score Bids

Weighted scoring across all criteria.

Each criterion carries a weight (e.g. Value for Money = 20%, ISO 27001 = 10%). The engine calculates weighted scores per bidder based on compliance status — compliant gates score higher, non-compliant gates score lower. Results are ranked to show the strongest submission.

5. Generate Report

A complete evaluation report with recommendation.

Collates everything into a structured report: procurement overview, criteria breakdown, per-bidder compliance matrices, deviation summaries, scoring rankings, and a final recommendation. Export-ready for audit trails and procurement panels.

Try it with our quick-test guide on the Evaluator page.

Go to Evaluator