Harvey AI: Harvey Legal AI Pricing, Who Harvey AI Law Tools Are For, and Alternatives
An independent look at what Harvey costs, what it does, and where it fits. If what you actually need is case law with a citation you can open, try the research box on the right first.
How courts have ruled
Sample results, illustrative only. Informational research, not legal advice. Verify every citation.
In short
Harvey is an enterprise legal AI platform sold to large law firms and corporate legal departments, covering research, drafting, contract review, bulk document analysis and agentic workflows. It publishes no price: as of 11 August 2026, harvey.ai/pricing returns a 404, there is no free tier, and there is no self-serve signup, so the only published way in is a demo request. Third-party estimates of its per-seat cost disagree with one another by more than tenfold, and reports of a roughly 20 seat minimum with a twelve month term are the reason most solo attorneys and small firms never get to a quote. If you need the full breadth of a firm-wide platform, Harvey is a serious product and its security posture is genuinely strong. If what you actually need is to find case law and cite it correctly, that is a much smaller purchase. Cases does that one job for a published 29 dollars per month, no quote and no minimum. Informational research, not legal advice, and always verify the citation.
Last updated August 2026
What is checkable, and what is only repeated
How much does Harvey AI cost
Search for Harvey pricing and you will find a dozen pages confidently naming a number. They do not agree with each other, which is the first clue. We checked the only thing that can actually be checked: whether Harvey publishes a rate. It does not. The URL harvey.ai/pricing returns HTTP 404, and no pricing page exists anywhere on the site.
So the table below splits the two kinds of claim rather than blending them. The published column is what Harvey itself states. The reported column is what third parties say, marked as exactly that. Most comparison pages merge these, and merging them is how an estimate quietly turns into a price.
| Question | Published by Harvey | Reported by third parties |
|---|---|---|
| Does Harvey publish a price? | No. harvey.ai/pricing returns HTTP 404. There is no pricing page on the site. | Not applicable, this one is directly checkable. |
| Per seat, per month | Nothing published. | Third-party write-ups put it anywhere from about 100 dollars per seat at Am Law scale to 1,200 or 2,000 dollars per seat for mid-sized firms. The estimates contradict each other by more than tenfold. |
| Seat minimum | Nothing published. | Commonly reported as a floor of roughly 20 seats, sometimes 20 to 50, with a 12 month term. Unconfirmed by Harvey. |
| Annual contract value | Nothing published. | Figures in the 50,000 to 200,000 dollar per year range circulate. Treat these as market gossip, not quotes. |
| Free trial or free tier | None offered. There is no self-serve signup. The only route in is "Request a Demo", which links to a contact-sales form. | No credible report of a public free trial. |
Published column read directly from harvey.ai on 11 August 2026, including the 404 on the pricing URL. Reported column summarizes third-party pricing write-ups and is presented as estimate, not price. Nothing here is a quote. Verify with Harvey before you budget.
The practical takeaway is not the number, it is the shape. A vendor with no pricing page, no free tier, no self-serve signup and a reported seat floor is telling you who it wants to sell to before you ever speak to anyone. That is a legitimate way to run a business, and it is also a fast way to work out whether you are in the market. Our legal research pricing breakdown sets Harvey against every other platform on the same published-versus-reported basis, and how much Harvey AI costs goes deeper on the estimates and where they come from.
The customer list is the answer
Who Harvey AI is actually for
You do not need inside information to work out Harvey's target buyer. Its own homepage names its customers, and the list reads: A&O Shearman, Dentons, Reed Smith, O'Melveny, Macfarlanes, PwC UK, Vinson & Elkins, BakerHostetler, Burges Salmon, CMS, Cuatrecasas, Gleiss Lutz, Schonherr, Thompson Hine, plus corporate legal departments at Comcast, Procter & Gamble, Deutsche Telekom, Merck KGaA, Repsol, KKR, Bridgewater, Syngenta and Adecco. Harvey also states that more than 70 Am Law 100 firms use it.
There is not a solo practitioner or a three attorney shop anywhere on that list, and that is not an oversight. Enterprise legal AI is sold through a procurement process because the deployment genuinely is one: security review, integration into a document management system, firm knowledge ingestion, admin controls, training. That work only makes sense amortized across a lot of lawyers.
Good fit
Firms with dozens to thousands of lawyers, a real procurement function, high matter volume, and a mandate to deploy AI across practice groups rather than for one task. If you have a legal ops team, you are in the market.
Awkward fit
Mid-sized firms of roughly 20 to 50 lawyers. Large enough to clear a reported seat minimum, small enough that the reported per-seat estimates land somewhere painful. This is where getting an actual quote matters most, because the public estimates are least reliable here.
Not the buyer
Solo attorneys, two to ten lawyer firms, paralegals and law students. A reported twenty seat floor makes the question moot before price enters it. Buy for the specific task instead, which is almost always research first.
Harvey has publicly signalled that it is beginning to look at smaller firms, nonprofits and government, so this may not hold forever. As of August 2026, though, nothing on its site is priced, packaged or self-serve for a small firm buyer, and a signal of future intent is not something you can put in a budget. If you are a solo or small firm evaluating options today, our guide to legal AI for lawyers sorts the market by task rather than by vendor, which gets you to a shortlist faster.
The modules Harvey names on its own site
What Harvey AI is used for
Harvey is a platform rather than a single tool, and the module names tell you how it is meant to be used: not for one job, but across a firm's work. Here is what it names, and what each part is for.
| Module | What it is for |
|---|---|
| Harvey Agents | Multi-step agentic workflows that carry out a legal task rather than answering a single prompt |
| Vault | Bulk document analysis across a large set of matter files |
| Knowledge | Answers grounded in the firm's own internal material alongside external sources |
| Shared Spaces | Collaborative workspaces so a matter team works from the same context |
| Command Center | Administration, deployment and oversight for the firm |
| Contract Intelligence | Contract review and analysis against firm standards |
| Harvey Mobile | Mobile access to the platform |
Module names as published on harvey.ai, August 2026. Descriptions are our summary of what each is presented as doing. Capability inside an enterprise contract varies by configuration, so confirm scope in writing.
Two of those, Vault and Knowledge, are the ones worth understanding because they explain the price. Both work over the firm's own documents: the matter file, the precedent bank, the internal know-how. That is a genuinely hard problem and it is not something a case law search tool does or claims to do. If your bottleneck is your own document set rather than published authority, that capability is the thing you are paying for and it is not available cheaply anywhere.
On what sits underneath, Harvey does not name its foundation models on its security page. It refers to "external model providers" in the plural, states that it contractually prohibits them from training on customer data, and requires Zero Data Retention from them. So it is a layer built over third-party models with contractual controls on top, which is true of most of this market and is not by itself a criticism.
Where Harvey is genuinely strong
Is Harvey AI confidential enough for client work
This is the part of Harvey we would not argue with. Its published security posture is what an enterprise buyer needs and it is more detailed than most of this market bothers to be: SOC 2 Type II and ISO 27001 audited annually, ISO 27701 and ISO 42001, GDPR and CCPA. It states plainly that it does not use inputs, outputs or uploaded documents to train underlying models, and that it requires Zero Data Retention from model providers. Hosting is Microsoft Azure, with processing available in the EU, Switzerland or Australia where data localization is required. Customer data is logically separated, and ethical wall policies can block restricted users from walled content.
Worth saying clearly: on this dimension Harvey beats us and beats most self-serve tools, and a firm with a real information security function will notice the difference immediately. We are a research tool you can sign up for in a minute, which is a different product with a different risk profile.
One caveat that applies to every vendor here, including us. ABA Formal Opinion 512, issued 29 July 2024, puts the duty on the lawyer to understand how a tool handles client confidences, and it says specifically that boilerplate consent buried in an engagement letter is not adequate. Certifications are evidence, not a substitute for reading the contract terms. We walk through what the opinion actually requires in can lawyers use AI for legal research.
Harvey wins three of these rows
Harvey AI alternatives, honestly compared
We sell a case law research tool, so read this table knowing that. We have tried to make it the version we would want to read if we were buying, which means conceding the rows Harvey wins rather than burying them.
| Dimension | Harvey | Cases | Honest verdict |
|---|---|---|---|
| Who it is built for | Large firms and corporate legal departments. Its own customer list is Am Law and Fortune 500 names. | Solo attorneys, small firms, paralegals and in-house lawyers who need case law fast. | Depends entirely on your seat count |
| Scope of work covered | Broad. Research, drafting, contract review, bulk document analysis, agentic workflows, firm knowledge. | Narrow on purpose. Case law research and precedent finding, nothing else. | Harvey, clearly |
| Pricing transparency | No published rate, no pricing page, quote only through sales. | 29 dollars per user per month, published, cancellable. | Cases |
| Time to first answer | Demo request, sales cycle, procurement, security review, rollout. | Type a question and read the cases. | Cases |
| Enterprise security posture | SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001, GDPR and CCPA, zero data retention required of model providers, data residency options. | Lighter. We are a self-serve research tool, not an enterprise deployment. | Harvey, clearly |
| Firm knowledge and your own documents | Yes. Vault and Knowledge are built for exactly this. | No. We search published US case law, not your document management system. | Harvey, clearly |
| Citator, is this case still good law | Not a published citator product. KeyCite and Shepard's remain the two established ones. | No citator either, and we say so on every page. | Neither. Use KeyCite or Shepard's |
If Harvey is out on seat count or budget, the useful move is to stop shopping for a platform and start shopping for the task that is costing you the most hours. For most solos and small firms that is case law research, and it is also the part of the market with real published pricing. Cases is 29 dollars per user per month and midpage is 30, both published and both self-serve. Paxton AI publishes 499 per user per month.
If you want a full research platform with agentic AI rather than a focused tool, Westlaw Advantage publishes 256.75 per user per month for a single circuit and 399.75 for all states plus federal, for firms up to ten attorneys, and CoCounsel Legal contains that same Advantage entitlement. Lexis+ with Protege and vLex Vincent AI are quote only, like Harvey. The full set sits on our legal research platforms comparison.
What buyers ask before the demo call
Harvey AI, answered
What is Harvey AI?
Harvey is an enterprise legal AI platform sold to large law firms and corporate legal departments. It covers research, drafting, contract review, bulk document analysis through its Vault module, and agentic workflows that carry out multi-step legal tasks. It is bought through a sales process rather than signed up for, and its own site lists customers such as A&O Shearman, Dentons, Reed Smith, PwC UK, Comcast and Procter & Gamble. Harvey states on its homepage that more than 70 Am Law 100 firms use it.
How much does Harvey AI cost?
Harvey publishes no price. As of 11 August 2026 harvey.ai/pricing returns a 404, there is no pricing page anywhere on the site, and the only route in is a demo request. Third-party estimates circulate widely but disagree with each other by more than tenfold, from roughly 100 dollars per seat per month at very large scale to 1,200 or 2,000 dollars for mid-sized firms. Anyone quoting you a single confident Harvey number is estimating, not reporting.
Is Harvey AI free?
No. There is no free tier, no free version and no published free trial. Harvey has no self-serve signup at all, so you cannot create an account and try it the way you can with a per-seat research tool. The only published path to access is the "Request a Demo" form, which routes to contact sales. If a free or immediate trial matters to you, that alone rules Harvey out.
Is Harvey AI good for small law firms?
Not today, in practical terms. Harvey is sold as an enterprise deployment with a sales cycle, and reports consistently describe seat minimums in the range of 20 seats with a twelve month term. A two lawyer firm cannot buy twenty seats. Harvey has publicly signalled interest in supporting smaller firms in future, but as of August 2026 nothing on its site is priced or packaged for a solo or small firm buyer.
What is Harvey AI used for?
The tasks its own site names are case and legal research, drafting, contract review and analysis, bulk analysis across a matter document set, working from the firm's internal knowledge, and running multi-step agentic workflows. In practice large firms deploy it as a general legal work platform across many matters rather than for one narrow job, which is part of why it is priced and sold the way it is.
What is Harvey AI built on?
Harvey does not name its underlying models on its security page. What it does state there is that it works with "external model providers" in the plural, that it contractually prohibits those providers from training on customer data, and that it requires Zero Data Retention from them. So it is a platform built over third-party foundation models with contractual controls layered on, rather than a single model of its own.
Is Harvey AI just a wrapper?
That framing is too blunt to be useful. Harvey does build on external foundation models, which it confirms on its own security page. But the products it sells are the layer above: bulk document analysis, firm knowledge retrieval, agentic multi-step workflows, admin and deployment controls, plus the security and residency commitments a large firm needs before any of it clears procurement. Whether that layer is worth an enterprise contract is a fair question. Whether it exists is not.
Does Harvey AI hallucinate?
Any system built on generative models can produce a citation that does not exist, and no vendor has eliminated this. The relevant question for a buyer is not whether it can happen but whether you can check quickly. That means asking whether the tool retrieves from an actual body of case law and shows you the source to open. The 2026 sanctions record is unambiguous that courts hold the filing lawyer responsible, not the software, so verification stays your job whichever platform you buy.
Is Harvey AI confidential?
Harvey publishes a substantial security posture: SOC 2 Type II and ISO 27001 audited annually, ISO 27701 and ISO 42001, GDPR and CCPA, and a statement that it does not use inputs, outputs or uploaded documents to train underlying models. It hosts in Microsoft Azure and offers processing in the EU, Switzerland or Australia for data localization. Separately, ABA Formal Opinion 512 puts the duty on you to understand how any tool handles client confidences, so read the contract terms rather than the badges.
Does Harvey AI have an API?
Harvey does not publish open API documentation the way a self-serve developer product does. Its site describes an Ecosystem alongside integrations into the tools firms already use. Integration scope is something worked out during the enterprise sales process, so if a specific integration is a requirement for you, put it in writing before signing rather than assuming it is covered.
What are the alternatives to Harvey AI?
It depends which part of Harvey you actually need. For case law research on a published per-seat rate, Cases is 29 dollars per month and midpage is 30. For a full research platform with agentic AI, Westlaw Advantage publishes 256.75 to 399.75 per user per month for firms up to ten attorneys, and CoCounsel Legal contains it. Paxton AI publishes 499. LexisNexis, Bloomberg Law and vLex are quote only, like Harvey.
Is Harvey AI worth it?
For a large firm with the seat count, the matter volume and a procurement function, the honest answer is that it can be, and the breadth is real. For everyone else the question rarely gets that far, because the seat minimum and the sales cycle decide it first. If your actual bottleneck is finding and citing case law rather than running a firm-wide platform, you are looking at a much smaller and much cheaper problem than the one Harvey is priced to solve.
Keep researching
Case law research without a sales call
Ask your research question in plain English and get on-point US federal and state precedents with the holding and a real citation. Published at 29 dollars per month, no quote, no seat minimum. Informational research, not legal advice, always verify the citation.