Introduction
Harvey is the legal AI platform most large firms end up looking at. It is sold to law firms, in-house legal teams and litigation practices, and the sale goes through a sales team. There is no self-serve sign-up. The homepage has one call to action, Request a Demo, and it shows no pricing and no trial.
The company has plenty of money. On September 9, 2026 Harvey announced a $550 million raise at a $15.5 billion valuation, co-led by Diffusion and Lightspeed Venture Partners. That was its second round of the year. A $200 million round in March valued it at $11 billion, according to Bloomberg Law. Harvey's homepage says more than 3,500 legal organizations and 200,000 professionals use the product across 85+ countries, and that 80+ AmLaw 100 firms are customers. An independent review from Haqq that we used for scoring puts AmLaw 100 adoption at 42%. Those two figures are far apart, and the reason is probably timing and definitions, so read both as approximate.
We gave Harvey a 3.6 out of 5. Its reach in Big Law is real. The score comes down because buyers can't see the pricing, there is no public API or documentation, and independent reviewers have reported fabricated case citations. The breakdown follows.
Key Features
Agents
Harvey describes Agents that carry out legal work end to end. Per the homepage, the system orchestrates complex matters and routes each task to the right agent. Harvey has also added a Legal Work Orchestration product (marked New on the site) for assigning and reviewing work. The vendor's pitch is that a lawyer hands over a multi-step task and reviews the result. We have not verified how well that holds up on real matters, and the independent reviews cited below are mixed.
Vault
Vault stores, organizes and bulk-analyzes legal documents. This is the feature that matters for due diligence, discovery review and any job where you need the same questions answered across a large document set. Users quoted by Ragbase say Vault has caps on what it can hold, and that gets in the way on large matters. Ask the sales team for the exact limits in writing before you sign.
Knowledge and research
Knowledge, per Harvey, answers legal, regulatory and tax research questions. Horizon Scanning tracks regulatory changes. This is the part of the product where citation accuracy matters most, so it is also where the reported hallucinations hurt most (see Cons).
Memory
Memory keeps user preferences across the product, so drafting style and working conventions don't have to be restated in every session. It was in the feature set we recorded when we scored Harvey in September 2026. The current homepage doesn't name it, so ask the vendor what it covers today.
Contract Intelligence
Contract Intelligence handles contract review and gives negotiation insights. For in-house teams this probably matters more than research, because most of their volume is commercial agreements.
Collaboration, intake and admin
The rest of the platform is built around firm operations:
- Spaces: shared workspaces for firms, clients and co-counsel.
- Intake: captures and routes incoming legal requests (new).
- Connector Library: security-reviewed connectors, plus custom MCP connections (new).
- Command Center: usage analytics and benchmarking, which suits firms worried about paying for seats nobody uses.
- Harvey Academy: on-demand training.
On security, Harvey lists SOC 2 Type II, ISO 27001, ISO 27701 and ISO 42001, GDPR and CCPA compliance, SAML SSO, audit logs and IP allow-listing. Large firm security reviews expect that list, and Harvey covers it.
Integrations are the weak spot. The Connector Library and MCP support are new. We found no public API reference or developer docs on the pages we checked or by search. If your plan is to build Harvey into your own document management or matter systems, expect that conversation to happen with an account team, with no docs to read first.
Pricing Breakdown
Harvey has no public pricing. Every plan is custom and quoted by sales.
| Plan | Published price | Third-party estimates |
|---|---|---|
| Custom | Contact sales | $1,200 to $2,000+ per seat per month; about $175,000 median annual contract (per Vaquill) |
Take those estimates as directional. Vaquill's analysis, the source behind our pricing score, puts seats at $1,200 to $2,000+ a month and the median annual contract at about $175,000. At those numbers a 20-lawyer team would pay roughly $288,000 to $480,000 a year before any discounts. Small firms are effectively priced out.
Opaque pricing creates a second problem. Ragbase and other reviewers report low seat utilization at some firms, which means firms are paying for seats that sit idle. Command Center exists to show usage. Before you commit firm-wide, negotiate a pilot with a small seat count and clear usage targets.
Scores
| Criterion | Score (out of 10) | Why |
|---|---|---|
| Output quality | 6 | Wide AmLaw 100 use and 38/50 on an independent benchmark (per Haqq), but citation hallucinations are reported |
| Ease of use | 3 | Users report workflows that don't match how they work, limited customization and vault caps (per Ragbase); no docs or self-serve trial |
| Pricing value | 2 | No public pricing; estimated $1,200 to $2,000+ per seat monthly (per Vaquill) |
| API and integration | 2 | No public API, docs or integration reference found |
| Problem fit | 5 | Agents and Vault target the right jobs, but reviews document fabricated case law and low seat use |
Overall rating: 3.6 / 5
Pros and Cons
Pros
- Big Law adoption. Haqq reports 42% of AmLaw 100 firms use it. Harvey's own site claims 80+. Either way, your opposing counsel and co-counsel may already be on it, and Spaces is built for working with them.
- Independent benchmark result. Harvey scored 38/50 on the independent benchmark cited by Haqq. That's a solid result, though well short of perfect.
- Broad coverage. One platform covers firm work, in-house work and litigation, from research through contract review and document analysis.
- Funding and staying power. With $550 million raised in September 2026 at a $15.5 billion valuation, Harvey is unlikely to disappear in the middle of your contract.
- Security paperwork. The certifications list will help it get through a firm's security review.
Cons
- No public pricing. Estimates run $1,200 to $2,000+ per seat per month. You can't model cost without a sales call.
- Reported fabricated citations. Ragbase and Haqq both report cases where Harvey produced case law that doesn't exist. In legal work that's a professional responsibility problem, so every citation needs human checking.
- Limited customization and vault caps. Users cited by Ragbase say the product doesn't flex to match their workflows, and Vault limits get in the way on big matters.
- No public API, docs or trial. Evaluating the product means a demo and a negotiated pilot. You can't read the docs ahead of time or try it yourself first.
Who Is It For
Harvey fits large law firms and large corporate legal departments that have a six-figure annual budget for legal AI, an innovation or knowledge management team that can run a rollout, and a review process that already checks every citation before it goes out. If your clients and peer firms are on Harvey, Spaces and shared workflows add to the case.
It fits poorly for solo practitioners, small and mid-size firms, and any team that wants to start with a self-serve trial. Engineering teams hoping to build on a documented API will also struggle. There is nothing public to build against, and the Connector Library is too new to judge.
Verdict
Harvey is the safe institutional choice for Big Law. Peer firms use it, the funding is there, it passes security reviews, and the product covers research, document review, contracts and multi-step agent work in one place. Those are good reasons for a large firm to shortlist it.
The 3.6 reflects what a buyer can't see. Pricing is hidden behind sales, and the estimates that do exist are steep. There are no public docs or API. Independent reviewers report fabricated citations, and some firms report idle seats. We recommend that a large firm with budget run a scoped pilot with a fixed seat count, written Vault limits and usage targets, and keep citation checking mandatory. If you are a smaller team, look elsewhere. You would be paying enterprise prices for a product sold and built around enterprise rollouts.