About
Market-review sheets tell you whether a tool is good. None of them tell you whether you can put a client’s return information in it. That is the question a WISP, IRC §7216, and the FTC Safeguards Rule actually ask.
- One verdict per tool, always tied to a use. Most land Conditional, and the conditions are the deliverable.
- Six posture dimensions, filled only from the vendor’s terms, privacy statement, DPA, or trust center. Never from a landing page.
- A scorecard and interop marks for API and MCP access, so a firm can see both the risk and the automation opportunity.
- General-purpose AI vendors included, with the consumer-versus-business tier split spelled out, because that is where members actually paste client data.
Honest status: shipped and in use. The methodology, the review queue, and the vetting rules are documented in the repo. Law and vendor terms change, so entries carry their verification date. It is owned by my firm, Charles J. Barmore CPA PC, and licensed to the AI Lab for hosting and display.
How I built it
- Hand-curated master, generated surfaces. A single JSON file is the source of truth. One Python script regenerates the typed TypeScript module, the full data export, and a gated HTML teaser.
- Batch vetting with an adversarial second pass. Tools were vetted in batches against primary sources, then re-judged. A cross-contamination bug caught in review is documented with its fix, and the grep that catches that class of error is written down.
- Released as a package so downstream apps pin versions and receive updates through a normal dependency bump.
Stack
JSON source of truthPython generatorTypeScript packageGitHub PackagesGitHub Actions