Paxton

What it is

Paxton is a US legal research and drafting assistant sold to individual attorneys and small-to-midsize firms, and it is the one tool in this catalog’s legal-research set that prints a per-seat number on its own pricing page. Four jobs run from one chat surface: research against a primary-law corpus, document drafting, analysis of uploaded files, and — added for plaintiff-side work — medical chronologies and medical billing summaries generated from raw treatment records. The company raised a $22M Series A led by Unusual Ventures, with Kyber Knight, 25Madison, and Wisconsin Valley Ventures joining.

The research corpus is US-only and appellate-and-above. Federal coverage runs to the United States Code, the Constitution, the Code of Federal Regulations, the Supreme Court, all thirteen circuits, district and bankruptcy decisions, and administrative bodies including the Tax Court, EEOC, PTAB, TTAB, and NLRB. State coverage runs to statutes, administrative codes, court rules, and supreme and appellate decisions across all 50 states. There are no secondary sources and no trial-court decisions. That boundary, not the model quality, is what decides whether Paxton fits a given practice.

Why it shows up in legal-ops stacks

• It is priced in public, and its peers are not. Every other legal-research assistant in this catalog — Harvey, Thomson Reuters CoCounsel, LexisNexis Protégé, vLex Vincent AI, Legora — is quote-gated. Paxton publishes a rate card and a 7-day free trial with no long-term contract, so a firm can price and test it in an afternoon instead of running a three-week demo cycle.
• The plaintiff-side record workflow is a shipped product. Medical Chronologies timestamps clinically significant events out of treatment records — accident dates, ER visits, lab results, therapies, complications — into an editable timeline, and Medical Billing Summaries does the same for billing files. This is the paralegal bottleneck in a personal-injury shop, and it is why Paxton’s own marketing has shifted toward PI, mass tort, and med-mal.
• It carries HIPAA alongside SOC 2 and ISO 27001. Feeding protected health information into a general legal assistant is a procurement problem. Paxton certifies against all three, with encryption in transit and at rest and quarterly access reviews.

Pricing reality

The Individual plan is $499 per user per month billed monthly, or $2,999 per user per year billed annually — about $250 per user per month effective, which is the 50% discount Paxton advertises. A 7-day free trial runs with no long-term contract. Enterprise is custom and volume-based, quoted against firm size and case volume, and adds firm-wide access, administrative controls, shared document sets, onboarding, and a named account manager.

The annual-versus-monthly split is the entire buying decision. At $499 monthly, Paxton sits above Westlaw Precision with CoCounsel, which a solo attorney with all-states-plus-federal coverage buys near $428 per user per month on a one-year term, and it sits at the top of the roughly $104–$639 band CoCounsel’s four standalone tiers span. At $250 effective on the annual commitment, Paxton undercuts that same Westlaw configuration by about 40% and the $639 Westlaw Advantage with CoCounsel Essentials bundle by about 60%. Paxton is not the cheap option month-to-month; it is the cheap option only against a twelve-month commitment. Firms that already hold a Westlaw subscription face a different sum entirely, because CoCounsel bolts onto it for roughly $100–$225 per user per month.

Best for

A solo or small-firm litigator — personal injury, employment, criminal, family, or regulatory — who wants research, drafting, and record review under one subscription and wants to evaluate it without a sales cycle. The published price and self-serve trial make it the practical default for a firm that has been quoted by two enterprise vendors and cannot get a number out of either.

Skip it if the firm runs on a practice-management system it expects the AI to read. Paxton does not integrate with Clio or any other LPMS, ships no public API, and runs no MCP server, so every matter document is a manual upload. Skip it as well if the work reaches outside the United States, or if it depends on trial-court decisions or secondary sources such as treatises and practice guides — none of that is in the corpus.

Versus the alternatives

The two incumbents to weigh it against are Thomson Reuters CoCounsel and LexisNexis Protégé. Pick CoCounsel when the firm already pays for Westlaw, because the add-on price beats a second full subscription and the research grounding is the corpus the firm already trusts. Pick Protégé when the firm is a Lexis shop, or when it needs the secondary sources and citator depth Paxton does not carry. Legora is the fastest-growing entrant in the segment and the pick when collaborative drafting across a multi-lawyer team matters more than a published price; Harvey is the pick at enterprise scale, where it now reaches primary law through its LexisNexis alliance. For plaintiff-side firms specifically, the record-workup rivals are Supio and EvenUp — pick Supio when medical-record throughput is the constraint and Westlaw grounding matters, EvenUp when the deliverable is a demand package backed by comparable verdicts, and Paxton when one subscription has to cover research and drafting as well as chronologies.

If none fit — a firm doing low-volume general litigation with a Westlaw subscription already in place — the cheaper move is the CoCounsel add-on on the existing contract rather than a second platform.

Watch-outs

• The accuracy numbers are vendor-run. Paxton reports a 94.7% non-hallucination rate and 93.82% accuracy on the Stanford legal hallucination benchmark, and 94% for its Citator on Stanford’s CaseHold. Both were run by Paxton, and a 2025 PNAS analysis of legal-AI benchmarking argues that parts of that benchmark are easy enough for simple bag-of-words models to score near-perfectly, so the figures do not transfer to hard research questions. Guard: during the 7-day trial, run ten questions you already know the answer to — including two where the controlling authority was recently overturned — and check every citation before the annual commitment.
• There is no integration surface at all. No LPMS connector, no public API, no MCP server. Every document enters by upload and every output leaves by copy-paste, which caps the throughput gain on high-volume matters and keeps work product outside the firm’s system of record. Guard: baseline the upload-and-retrieve time on one real matter during the trial, and price the annual seat against that measured number rather than the demo.
• Communications with Paxton are not privileged. Paxton’s own disclaimer states that exchanges are covered by its privacy policy but not by attorney-client privilege or the work-product doctrine, and the medical modules mean PHI is in scope. Guard: sign a BAA before any client records go in, confirm in writing whether inputs train models and how long they are retained, and set a firm policy on what may be pasted into the assistant.
• The monthly plan is a trap for a slow evaluation. Three months of month-to-month billing costs $1,497, half the annual seat, with nothing carried over. Guard: treat the 7-day trial as the real evaluation window, and move to annual or drop the tool rather than idling on monthly.