Harvey
Also known as: Harvey AI Corporation
Enterprise legal AI platform with planned and approved agents, Vault document analysis, licensed research, memory, an agent builder and a connector library, used by more than 3,500 legal organizations.
Harvey is an enterprise legal AI platform for law firms and in house legal teams, built by Harvey AI Corporation of San Francisco. It says more than 3,500 legal organizations and 200,000 professionals in over 85 countries use it, including more than 80 of the Am Law 100. Harvey raised $550 million at a $15.5 billion valuation in September 2026, co-led by Diffusion and Lightspeed Venture Partners, and agreed in October to acquire Aida, its fifth acquisition of the year. Cozen O'Connor, with more than 1,000 attorneys, and Spencer West chose Harvey firmwide in November 2026, and CVS Health is a customer.
Its agents take a legal task from description to a cited, review ready deliverable. The user previews and approves the plan before work begins, Harvey coordinates several agents in parallel and brings the user back when judgment is needed, and every step is logged with a citation on every claim.
Vault holds up to 100,000 documents per vault for bulk analysis, Knowledge researches across more than 500 licensed sources such as LexisNexis and Wolters Kluwer, and Harvey Memory keeps each user's preferences across sessions, reviewable and deletable.
More than 500 prebuilt agents cover practice areas, and Agent Builder, Workflow Builder and Legal Work Orchestration, launched in October 2026, let firms encode their own processes.
The Connector Library, generally available since October 21, 2026, links Harvey to iManage, NetDocuments, SharePoint, Google Drive, Box and dozens more tools, and the Harvey MCP server lets Claude, Gemini Enterprise and Microsoft 365 Copilot query Harvey. Admins can let users pick models from Anthropic, OpenAI and Google, with automatic routing by default.
Harvey holds SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001 and AIUC-1 and announced FedRAMP Class D in October 2026. It runs on Microsoft Azure, keeps data in the EU or Switzerland, the US or Australia as the customer chooses, and supports SAML sign on, ethical walls and customer set retention. Pricing is by enterprise contract, with no published rates.
Vendor details
Canonical URL
https://harvey.ai
Category
Enterprise operations agent
Subcategory
Legal — enterprise legal AI platform
Funding status
Venture backed. Harvey raised $550 million at a $15.5 billion valuation on September 9, 2026, co-led by Diffusion and Lightspeed Venture Partners. Harvey AI Corporation is based in San Francisco.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
The Connector Library, generally available since October 21, 2026, connects Harvey to document management, data room, research and collaboration tools including iManage, NetDocuments, SharePoint, Google Drive, Box, PitchBook, Intralinks, Slack and Docusign, with write back to document management systems and custom MCP connections. The Harvey MCP server lets Claude, Gemini Enterprise and Microsoft 365 Copilot query Harvey, and APIs cover completion, Vault, intake, client matters, audit logs and history export.
In practice
Due diligence means a junior associate reading thousands of contracts by hand. Harvey's Vault holds up to 100,000 documents per vault and its agents review across them, so the first pass isn't all human hours.
Your firm has a repeatable workflow you wish ran itself. Harvey's Agent Builder lets teams encode it as a custom agent that runs in Harvey and in Word, alongside more than 500 prebuilt agents across practice areas.
A complex matter needs many steps, not a single answer. Harvey plans the work for the lawyer to approve, splits it across agents running in parallel and asks the lawyer in when judgment is needed, returning a cited deliverable.
Agentic Index coverage score
12.5 / 14 capabilities · 89%
| Integrations & Tool Calling | Full |
|---|---|
|
The Connector Library, generally available since October 21, 2026, gives agents security reviewed connectors to the team's tools, with write back starting with document management systems, and customers can add their own MCP connections. Its page names dozens of connectors, including iManage, NetDocuments, SharePoint, Google Drive, PitchBook, Intralinks, Box, Slack and Docusign, and agentic search reaches across iManage, SharePoint and Google Drive. Sourceharvey.ai/blog/harvey-connector-libraryread 2026-11-03 |
|
| Workflow Orchestration | Full |
|
Harvey coordinates a team of agents that split large tasks, work in parallel and return review ready outputs, plans each task before running it and carries legal work from description to deliverable. Legal Work Orchestration, launched October 15, 2026, routes matters through steps with notifications, review gates and email replies, and Agent Builder and Workflow Builder let customers define their own agents of several steps, which run in Word since September 2026. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Knowledge Grounding & RAG | Full |
|
Vault stores, organizes and analyzes up to 100,000 of the customer's documents per vault as persistent projects queried with source citations, and it syncs with iManage, SharePoint and Google Drive. Knowledge grounds research in more than 500 licensed legal sources, including LexisNexis, Wolters Kluwer and Lefebvre, and Vault projects and knowledge sources are addressable through the API and MCP server. Sourceharvey.ai/platform/vaultread 2026-11-03 |
|
| Human Oversight & Guardrails | Full |
|
Before a run, the user previews the plan, adjusts the scope and approves the work before Harvey begins, and Nudges bring the user in mid run when judgment is needed before work continues. Legal Work Orchestration adds review gates between steps, and admins can require approval before a custom agent is shared at the levels they choose. Output review then happens in the user's Word document. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Security, Identity & Governance | Full |
|
Certifications include SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001 and AIUC-1, and Harvey announced FedRAMP Class D on October 28, 2026, with TX-RAMP Level 2 and IRAP listed on its trust center at trust.harvey.ai. Customers get SAML single sign on, SCIM sync through Command Center, IP allow listing, retention policies they set with deletion at any time, and their own ethical walls kept in sync, now generally available for iManage. Workspace admins manage per user permissions and feature gates, and Harvey requires zero data retention from its model providers. Sourceharvey.ai/securityread 2026-11-03 |
|
| Observability & Auditability | Full |
|
Every claim in agent output is backed by a citation and every step is logged for verification before partner or client review. Audit logs are among the standard controls, and the API offers audit log and history export endpoints so a firm can pull the record into its own systems. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Memory & State Persistence | Full |
|
Harvey Memory saves preferences in the flow of work, personal to the user, and keeps them across sessions and surfaces, from the web app and Outlook to the Word add in and playbook runs. They are reviewed, edited or deleted from one place and granted by role through admin controls. Agent Builder configuration is a separate asset. SourceHarvey, harvey.ai/platform/memoryread 2026-09-06 |
|
| Deployment & Data Residency | Full |
|
The customer keeps everything in region, choosing the EU or Switzerland, the US, or Australia, and the choice extends to Harvey's subprocessors. Harvey hosts its cloud environment in Microsoft Azure, with separate US, EU and Australian application and API hosts. Harvey is hosted only and does not describe a VPC or on premises option. Sourceharvey.ai/securityread 2026-11-03 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Harvey ships prebuilt use case agents a customer deploys as they are, from regulatory briefing and deposition preparation to deal point analysis, precedent drafting and case strategy, and it announced more than 500 of them across practice areas in May 2026. Legal Work Orchestration adds workflow templates in early access, and Agent Builder is the customer's own tooling for building more. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Triggers & Channel Coverage | Full |
|
Scheduled Agents run reports, updates and reviews automatically on a schedule the customer chooses, and Legal Work Orchestration moves work forward on notifications and email replies. Users reach agents from the web app, the Word add in, Outlook and Harvey Mobile, and from outside assistants such as Claude through the Harvey MCP server. Harvey does not describe webhook or event subscriptions. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Model Flexibility & Routing | Full |
|
Harvey routes automatically by default, and workspace admins can switch on a Model Selector that lets users pick the model in Assistant, Vault and Workflow Builder. October 2026 releases added Claude Opus 5.5 and Sonnet 5.5, new GPT models and Gemini 4 Argon to the selector, so the choice spans Anthropic, OpenAI and Google. Sourcehelp.harvey.ai/release-notesread 2026-11-03 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
The Harvey API at developers.harvey.ai offers an Open Ended completion endpoint on Harvey's own models with bearer token authentication and an EU hosted API host, sold as an extra through the account manager. Alongside it, a remote Harvey MCP server over Streamable HTTP with OAuth exposes ask, Vault query and knowledge source tools to Claude, Gemini and Microsoft 365 Copilot, scoped to each user's permissions and switched on per user by the admin. SourceHarvey, developers.harvey.ai/guides/harvey_mcp and developers.harvey.ai/reference/completionread 2026-09-06 |
|
| Testing, Debugging & Optimization | Partial |
|
Command Center gives administrators usage analytics and peer benchmarking on the organization's use of Harvey, and Harvey publishes its own Legal Agent Benchmark, LAB, as an open source project. Harvey does not describe a test harness, test cases or release gate customers can use for agents built in Agent Builder or Workflow Builder. Sourceharvey.ai/platform/agentsread 2026-11-03 |
|
| Browser & Computer Use | Not documented |
|
Agents reach systems through connectors, MCP and the API, and Harvey runs as a web app, a Word add in, an Outlook integration and a mobile app, all its own surfaces. Harvey does not describe browser, desktop or computer control. SourceHarvey, harvey.ai/platform/agentsread 2026-09-06 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
BYO (Custom MCP) lets an organization register its own MCP servers and connect any MCP compatible tool to Harvey. Workspace admins decide which tools are connected and who can use them.
Bears on: MCP / tool calling / API
View sourceLegal Work Orchestration organizes a matter in Harvey into phases and tasks, each carrying its scope, inputs and reviewer, so agents can execute defined tasks and lawyers review the output at designated gates. Harvey notifies the reviewer when an agent finishes, moves to the next step on approval, and lets lawyers close or start tasks by replying to email, while reusable matter templates are in early access.
Bears on: Human approval / guardrails
View sourceA new Connector Library in Harvey is a catalog of admin enabled connectors that are either native integrations or vendor run MCP servers its agents call as tools. Harvey says each MCP tool is security reviewed first, every call passes a policy engine that fails closed, and admins can switch off a whole connector or a single tool.
Bears on: Security / enterprise
View sourcePricing
Contact for pricing
What is public
Harvey publishes no pricing page and no rate card, and the site offers Request a Demo only. API completion access is an additional purchase arranged through the account manager.
Billing mechanics
Not disclosed. Enterprise contracts are quoted through sales.
Cost watchouts
API completion access is an additional purchase through the account manager. No other commercial terms are published.
Additional watchouts
There is no public pricing, no free tier and no self serve path. The API is an additional purchase.
Sales call required
Yes, required for paid access
Free / trial
None published, demo on request
Lowest paid plan
Not published
Key ambiguities
No official price is published.
Support SLA / resale
Enterprise onboarding, dedicated solutions and onboarding teams, and enterprise security and compliance come standard. The specifics are set by contract.
Missing data
There is no pricing page. Seat price, seat minimums, term and onboarding fees are not published.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Adopt AI — Finance agents that categorize, reconcile and close the books inside…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
Alternatives to Harvey
The closest documented capability profiles to Harvey among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Dust12.5 / 14Matches Harvey across all 14 documented capabilities
- Glean12.0 / 14A lighter documented profile than Harvey
- Salesforce13.0 / 14Fuller documented coverage on Testing, Debugging & Optimization
- Workato12.0 / 14A lighter documented profile than Harvey
- Atlassian Rovo13.5 / 14Adds documented Browser & Computer Use
- Boomi12.5 / 14Fuller documented coverage on Testing, Debugging & Optimization
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded