This system handles the documentation pipeline for Canadian SR&ED tax credit claims by ingesting project artifacts and running them through structured eligibility scoring before generating T661 narratives. You upload PDFs, Jira exports, code diffs, or logs into an IDE-style workspace where everything stays visible on one screen. The system indexes these files asynchronously and builds a searchable evidence base that backs every claim with traceable citations.
The core workflow starts with evidence ingestion. Files get processed through an indexing pipeline that extracts content and metadata. Once indexed, you can run a structured eligibility pre-check that scores your work against the three SR&ED pillars. This isn't a binary yes-no scan. It breaks down technical uncertainty, systematic investigation, and technological advancement into separate scores with specific evidence gaps flagged.
After eligibility checks, the system generates T661 draft narratives. These aren't generic descriptions. Every statement in the output links back to specific source documents through a citation system. You see which PDF page or Jira ticket supports each claim. This evidence trail matters for audits because CRA reviewers can trace assertions to original artifacts without hunting through file folders.
The workspace includes batch processing for multiple projects and an API interface for programmatic access. Time extraction outputs CSV files that map hours to specific activities. The logbook generation pulls together chronological records of experiments and iterations. Risk scanning identifies weak spots in your claim where evidence looks thin or methodology appears unclear.
Technical limitations are straightforward. The free plan caps you at one project with five files and restricts AI workspace runs. That's enough for initial eligibility checks but not full claim preparation. The Starter plan at 149 CAD monthly gives you five projects, 25 files per project, and 50 AI runs per month. Professional tier at 399 CAD monthly removes project limits and pushes file counts to 100 per project with unlimited AI runs. Enterprise pricing is custom.
File count caps matter more than they might seem. A typical SR&ED claim pulls from sprint retrospectives, architecture docs, failed experiment logs, and technical specifications. Five files cover basic eligibility checks. Twenty-five files handle small team claims. One hundred files support larger organizations with multiple concurrent R&D streams.
The system integrates with Jira through export support. You export your Jira data and upload it rather than connecting via live API. This means you're working with snapshots, not real-time feeds. No other integrations are documented.
Output quality depends entirely on input quality. The citation system prevents hallucination by tying every claim to source material, but it can't create evidence that doesn't exist. If your documentation has gaps, the system identifies them but won't fabricate details to fill holes. This is a feature for audit compliance but means you can't generate strong narratives from weak documentation.
The evidence-first approach differs from traditional claim preparation where consultants interview teams and reconstruct timelines from memory. Here, everything starts with existing artifacts. That makes the process faster but requires decent documentation habits during development. Teams that don't log experiments or track failed approaches won't have much to feed the system.
An optional Expert Review add-on costs 399 CAD one-time and brings human review to generated drafts. This connects automated output with submission-ready claims.
No guarantee of CRA compliance is mentioned anywhere. The system produces audit-friendly documentation with traceable sources, but acceptance depends on the underlying work meeting SR&ED criteria and your evidence supporting that case convincingly.
