CoCounsel vs Harvey. One arrives through your research contract, the other through a demo request.

CoCounsel has the longer lineage of the two. Thomson Reuters completed its acquisition of Casetext on August 17, 2023 for $650 million in cash; at the time Casetext described more than 10,000 law firm and corporate legal department customers and a CoCounsel product running on GPT-4. Thomson Reuters then spent two years pulling that assistant into its own stack, and in August 2025 launched CoCounsel Legal alongside Westlaw Advantage. By February 2026 Thomson Reuters reported one million CoCounsel users across 107 countries, and said the product works with frontier models from Anthropic, OpenAI and Google alongside its own technology, with customer inputs not used to train third-party models.

What that buys a firm is grounding. Deep Research plans a multi-step research approach, executes it, and returns a report cited back to Westlaw and Practical Law, with a verification pass that checks whether the cited authority actually supports the assertion. For litigation and research-led practices this is the shortest distance between a question and a defensible answer, and it is built on content the firm's associates already trust. Thomson Reuters also signalled that Westlaw Advantage is likely the final versioned release of Westlaw, with improvements arriving as continuous updates rather than a new product every four years. For a firm that has lived through Westlaw Next, Edge and Precision, that is a real procurement fact and not just marketing.

Harvey is the better-funded and faster-moving of the two as an independent company. In March 2026 it announced $200 million at an $11 billion valuation co-led by GIC and Sequoia, past $1 billion raised in total, with more than 100,000 lawyers across more than 1,300 organizations in over 60 countries, the majority of the AmLaw 100 among them, plus more than 500 in-house legal teams. Named firmwide wins in that announcement include DLA Piper International, Corrs Chambers Westgarth and McCann FitzGerald, with NBCUniversal and HSBC on the in-house side.

The product reflects who bought it. Harvey describes itself as an operating system for legal and professional services and says it runs more than 25,000 custom agents across M&A, due diligence, contract drafting and document review. Vault, Knowledge, Shared Spaces and Command Center are the surface of a platform designed for large matters with many people touching them. If your firm's economics are driven by a handful of large deals or large-document matters rather than by research volume, Harvey is built closer to your work than CoCounsel is, and the reference base at that shape of work is deeper.

Most comparisons of these two treat them as competing AI products. They are also, and more consequentially, competing distribution arms for rival research libraries. Thomson Reuters owns Westlaw and Practical Law and built the assistant on top. Harvey owned the assistant and, in June 2025, signed an alliance with LexisNexis to bring in primary law content and Shepard's Citations, with jointly developed workflows for motions to dismiss and motions for summary judgment named in the announcement.

That makes the buy partly a bet on which library your firm wants to be deeper inside five years from now. Thomson Reuters markets CoCounsel Legal as the only legal AI reasoning from authoritative Westlaw primary law and trusted Practical Law guidance. Harvey's Lexis alliance is the direct answer to that claim, and the practical result is that both roads now end in a citation-backed report grounded in a major library. The differentiator that survives is not whether the AI can cite. It is whose citator your associates already know how to challenge.

The mid-market consequence is unglamorous: switching research libraries to follow an AI product is almost always the more expensive move, because the cost is retraining and habit rather than licence fee. Firms that end up unhappy usually got there by buying the AI first and discovering the library commitment second.

Here is the whole of what either vendor publishes. Thomson Reuters lists Westlaw Advantage from $256.75 per month for single-circuit coverage and $399.75 per month for all states and federal, with discounts of 12 percent on a two-year term and 18 percent on three, and states that online pricing is available only for firms with up to 10 attorneys. Its CoCounsel Legal plans page names four plans and publishes no dollar figure. Harvey publishes no figure at all.

That means the only published price in this comparison is for a firm smaller than the reader of this page, and it is for the research library rather than for the AI. Everything else in circulation is third-party estimation. One frequently cited analysis reports roughly $1,000 to $2,000 per user per month for firms of 50 to 200 attorneys and roughly $100 to $200 per user per month at AmLaw 100 scale, with seat minimums around 25 to 50 and twelve-month commitments, and states outright that the numbers are triangulated from industry reporting and customer disclosures rather than published by the vendor. Another reports CoCounsel at roughly $100 to $200 per user per month as an add-on rising to $300 to $600 or more once Westlaw is included, again from firm reports and analyst estimates.

We are repeating those ranges because a buyer will find them anyway, and it is better to see them labelled honestly. They are reported, not confirmed. They disagree with each other by an order of magnitude at mid-market seat counts. Do not build a board paper on them. Run both quotes simultaneously, ask each vendor for the per-attorney number at your exact headcount and term, and hold the two quotes next to the fixed-fee commission bands before anyone signs.

The mistake we see most often is a firm asking whether a custom build can replace CoCounsel or Harvey. It cannot, and it should not try. Neither can be beaten at grounded research by a boutique that does not own a research library, and pretending otherwise would be the fastest way to lose a client's trust in month three.

What a commission does is different in kind. It takes one workflow that is specific to your firm, that no vendor can represent because no vendor has your data, and builds a system around it: intake to matter routing on your taxonomy, conflicts against your own history, billing narrative reconciliation, referral attribution, precedent retrieval from your own archive with your own permission model intact. The fee is fixed at $45,000 to $180,000 in two installments, one at production-build start and one at handoff. The code, prompts, models and data pipelines transfer to the firm at handoff and run inside the firm's own cloud tenant. There is no per-seat line and no renewal.

The healthy end state for most mid-market firms is two systems, not one: a product for legal work product, a commissioned build for the operational middle, and a clean boundary between them. Firms arrive at that shape faster when they buy the product first and commission second, once they can point at exactly what the product did not reach. Our law-firm diagnostic exists to find that boundary in one call.

Thomson Reuters publishes four plan shapes and three of them are named after a research product: CoCounsel Legal, Westlaw Advantage with CoCounsel Essentials, Practical Law Dynamic Tool Set with CoCounsel Essentials, and CoCounsel Essentials. The Essentials plan is described as document analysis and drafting from the web to Microsoft Word, without the research library attached. So a firm can buy the assistant on its own. But the capability Thomson Reuters markets hardest, Deep Research grounded in Westlaw and Practical Law, is the one that arrives with the research contract.

Harvey publishes a solutions page for boutique, specialty and mid-sized firms, so the answer is not no on principle. What Harvey does not publish anywhere on its site is a price, a seat minimum, or an attorney count. Every figure circulating for mid-market Harvey deployments is third-party reporting, and those reports disagree with each other by roughly a factor of ten. Treat Harvey as a real option and treat every number you read about Harvey as unverified until a quote arrives in writing.

They are grounded in rival research libraries, so the honest answer depends on which library the firm already trusts. CoCounsel Legal reasons over Westlaw and Practical Law, both owned by Thomson Reuters, and returns citation-backed reports against that content. Harvey announced an alliance with LexisNexis in June 2025 under which Harvey customers can select the LexisNexis Protege service and receive answers grounded in LexisNexis case law and statutes, validated through Shepard's Citations. A firm that cite-checks in KeyCite and a firm that cite-checks in Shepard's are not asking the same question.

Yes, and some firms do, usually because the research renewal brought CoCounsel in and a practice group bought Harvey for deal work. It is expensive, and it splits the firm's institutional knowledge across two systems that do not talk to each other. A firm that is going to run both should decide the boundary before the next renewal rather than after: which practice groups use which, which system holds the precedent library, and what happens to the work product inside whichever one eventually gets cut.

When the constraint is not legal work product. Both vendors are shaped around research, drafting, review and diligence, and both are good at it. Neither is shaped around routing an intake into the right matter type on a taxonomy only your firm uses, running conflicts against your own matter history, or reconciling billing narratives against what actually happened on the matter. A commissioned build is the right answer when the firm can name that workflow in one sentence, has the data to build it on, and wants to own the result. It is the wrong answer when the leak is research hours, because that is the problem the products already solve.