Drop a .docx. BarZero runs clause-level analysis against your preferred positions and fallbacks, flags deviations, and writes the redlines back as native Word tracked changes. The original document is parsed in your browser and never uploaded in its binary form.
Indemnity, limitation of liability, IP assignment, termination, auto-renewal, exclusivity, and the rest of the usual suspects — each clause scored against your playbook rather than a generic risk list.
Your preferred clauses and fallback positions live in a browser-local library. Only the specific entries relevant to the contract under review are injected into the prompt — the library itself never leaves your device.
Every clause is compared against your preferred position and your fallback. BarZero labels the gap — deviation, partial match, or clean — and proposes redline language from your playbook, not a generic template.
Click Export redlined .docx and BarZero injects every flagged redline into your original document as native Word tracked changes — author shows up as ‘BarZero’, you accept or reject each one in Word the same way you would for a human reviewer. The writer runs entirely in your browser.
When you open a document in BarZero, the .docx is parsed in your browser with mammoth. Author metadata, comment threads, and tracked-change history are stripped locally with jszipbefore anything else happens. Only the plain text — plus the specific playbook entries relevant to the clauses we're reviewing — is relayed over TLS to a stateless handler that forwards to Claude under Anthropic's Zero Data Retention agreement.
The handler holds the request in memory long enough to stream a response back, then returns. Nothing is persisted server-side — no prompt log, no document store, no cache. Because contract review uses no tool connectors, ZDR is active unconditionally.
Point BarZero at a folder of contracts and review every one with a single playbook position. Each contract streams through the same zero-retention inference path as a single-doc review, and when the run finishes BarZero writes an aggregate index that groups every contract by headline finding (red / amber / clean) and lists the top cross-contract risks.
Batch review is available on Ensemble and higher — see pricing for details.
ABA Model Rule 1.6(c) requires attorneys to take reasonable steps to prevent unauthorized disclosure. BarZero parses the document in your browser, strips metadata locally, sends only plain text under ZDR, and retains nothing server-side. Consistent with the Rule and Formal Opinion 512 on good-faith use of generative AI.