Solutions

Review a structured matter before you review a draft

Patendra keeps retrieved sources, claim versions, review notes, and draft sections connected. Start from a visible evidence record, inspect what the software could not resolve, and apply professional judgment without reconstructing the work across separate tools.

§102 · §103 · §112 per-statute memos ID-verified citations Local-first Attorney review required
A patent attorney annotating a printed claims page with a fine pen

For your practice

A more structured first pass

It does not replace prosecution judgment. It makes the hours before your judgment applies cheaper and better documented.

Citations

Evidence-gated citations you can verify by ID

Every reference behind a simulated §102 or §103 rejection comes from hybrid retrieval across USPTO Open Data, Espacenet, Google Patents BigQuery, and a local vector store, fused with Reciprocal Rank Fusion. A citation only counts if it was actually retrieved and carries a verifiable document ID, so you can pull the reference yourself before you rely on anything.

Memos

Per-statute memos that read like Office actions

Findings arrive organized the way you already think: anticipation under §102 with element-by-element mapping, §103 combinations with an articulated motivation to combine, and §112 written-description, enablement, and definiteness issues, each in its own memo rather than blended into one verdict.

Drafts

Drafts that arrive pre-examined

The specification, claim set, and Graphviz figures you receive were refined through a three-tier best-first search and a simulated examination before you ever see them. It is a starting point for your drafting, not a filing, but one that has already survived an adversarial pass.

Record

A file-wrapper-friendly artifact trail

Every run preserves its intermediate artifacts (search logs, retrieved references, examination memos, score tables, and the claim-refinement history) so you can reconstruct why a claim was broadened or narrowed and fold the search record into your own diligence.

And because Patendra is a local-first desktop app for macOS and Windows, client disclosures never leave the machine except for calls to the LLM provider you configure with your own Anthropic Claude or Google Gemini key. Details in the security model.

Honesty features

Built for professional skepticism

You've seen tools that are confidently wrong. Patendra is designed so its confidence is always inspectable and bounded by what the search actually found.

No verdict theater

No single "ALLOWABLE" badge

Results are reported per statute, always. A claim set can look clean under §102 and still have a live §103 combination or a §112(b) definiteness problem. The dashboard never collapses that into one green light.

Search honesty

Insufficient-search warnings

"No anticipating art found" is only meaningful if the search had real coverage. When prior-art retrieval was too thin to support a novelty conclusion, the report says so explicitly instead of letting silence read as allowance.

Reproducibility

Deterministic scores

Novelty, non-obviousness, enablement, and breadth are computed from measurable signals in the evidence record: the same references and findings always produce the same score. The model never gets to make up the number.

Clear labeling

Qualitative analysis, labeled as such

The raw qualitative LLM analysis is preserved and shown, clearly labeled "for review, not scoring." You get the model's reasoning as context for your own, without it ever contaminating the deterministic numbers.

Workflow fit

Where it sits in your matter workflow

Patendra slots in between intake and your own drafting pass. The full workflow is described on how it works.

  1. Intake the disclosure

    Enter the invention disclosure as it came from the client, locally on your machine. Nothing is uploaded to us.

  2. Run the workflow

    Ideation, hybrid prior-art retrieval, three-tier claim refinement, and simulated examination run end to end, streaming progress as they go.

  3. Review the memo, statute by statute

    Read the §102, §103, and §112 findings the way you'd read an Office action, with every asserted reference retrievable by ID for your own verification.

  4. Record your assessment

    Rate the run Strong, Moderate, Weak, or Rejected. Your verdicts feed the local calibration loop, so scoring confidence adjusts to your practice's standards over time.

  5. Refine and file with your judgment

    Take the draft and the evidence package into your own drafting and prosecution strategy. The attorney-review banner on every output means what it says: the filing decision, and the filing quality, are yours.

Professional responsibility. Patendra is a drafting and search aid, not counsel and not legal advice. Duties of competence and candor, and confidentiality review (including reviewing the terms of the LLM provider whose API key you configure) remain with you. Every output carries an attorney-review banner for a reason.

Review it with your most skeptical partner

Practice plans include five seats, a shared self-host server, and portfolio runs, with your clients' disclosures staying on your infrastructure.

See pricing