Every document goes through the same structured process: upload, configure, process through a multi-agent pipeline, and download the approved result. Here is exactly what happens at each stage.
Drop in a markdown file up to 50,000 words. The system detects the document type automatically and recommends analysis types based on the content. Choose Revise Document to improve your draft, or Generate Analysis to produce a structured critique.
Select from 27 specialized analysis types or accept the system recommendation. Review the estimated cost, token count, and processing time. Set your confidence threshold -- the minimum score the pipeline must achieve before output is approved.
The pipeline runs through Writer, Reviewer, and Reconciler stages. If the confidence score falls below your threshold, the system automatically cycles through additional revision rounds. You can monitor progress in real time.
Once the pipeline meets your confidence threshold, download the approved document as markdown or HTML. The output includes the analysis, the confidence score, and a summary of what each pipeline stage contributed.
The core of AnalyzeMyDocument is a multi-agent pipeline where each stage has a distinct role. No single agent controls the entire output.
The Writer agent produces the initial analysis or revision based on the selected analysis type. It receives the original document, the analysis type prompt, and any configuration parameters. The Writer's output is a complete draft -- not a final product.
Produces the initial draft
Three independent Reviewer agents examine the Writer's output against the original document. Each Reviewer operates with a different critique perspective and produces structured feedback: issues found, severity ratings, and specific suggestions. Reviewers do not see each other's output.
Three independent perspectives
The Reconciler agent receives all three Reviewer reports and the Writer's draft. It resolves conflicts between reviewers, weighs the severity of each issue, and produces a final confidence score (0.0 to 1.0). If the score meets the threshold, the output is approved. If not, the Reconciler sends structured feedback to the Writer for the next cycle.
Resolves conflicts and scores confidence
The confidence score is not a vague quality rating. It is a structured assessment produced by the Reconciler after evaluating all three Reviewer reports.
The Reconciler evaluates the Writer's output across multiple dimensions: completeness, accuracy, consistency with the original document, alignment with the analysis type requirements, and resolution of Reviewer-flagged issues. Each dimension contributes to the final score on a 0.0 to 1.0 scale.
0.0 -- 0.3: Major issues remain. The output has significant gaps or inaccuracies.
0.3 -- 0.5: Partial coverage. The output addresses some concerns but misses important elements.
0.5 -- 0.7: Solid analysis. Most issues are covered with reasonable depth.
0.7 -- 1.0: Thorough analysis. Comprehensive coverage with strong supporting reasoning.
You set the minimum confidence threshold before a run starts. The pipeline will cycle through additional revision rounds until the score meets your threshold or the maximum cycle count is reached. Higher thresholds produce more thorough output but take longer and cost more in processing time.
The final output includes the confidence score, a breakdown of how it was calculated, and a summary of what each Reviewer flagged. You can see exactly why the pipeline approved the output at the score it reached.
Unlike single-pass AI generation, AnalyzeMyDocument iterates. If the first pass does not meet your confidence threshold, the pipeline revises and re-evaluates.
After the Reconciler scores the output, it checks against your confidence threshold. If the score is below the threshold, the Reconciler generates structured feedback for the Writer, identifying exactly what needs improvement. The Writer produces a revised draft, the Reviewers evaluate it again, and the Reconciler re-scores. This continues until the threshold is met or the maximum cycle count is reached.
Free: 1 cycle (single pass).
Quick Review ($14.99): Up to 2 cycles.
Standard ($29.99): Up to 3 cycles.
Deep Analysis ($49.99): Up to 5 cycles.
If the pipeline exhausts all available cycles without meeting your threshold, you still receive the best output produced. The final score and Reconciler notes are included so you can see where the analysis fell short and decide whether the output is still useful for your purposes.