# Harvey

Commercial AI platform for legal work with agents, document analysis and shared matter workspaces.

Canonical profile: https://ki-agentenkompass.de/en/agents/harvey/
Editor: [Michael Meierhoff](https://ki-agentenkompass.de/en/about/)
Sources reviewed: 2026-11-05
Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests.

## Overview

Harvey is a proprietary, sales-led product. Agents are one part of the platform alongside Vault, Spaces and Memory.

- Provider: Harvey
- Product family: Harvey
- Profile type: Legal AI platform with agents
- Environment: Web app; according to the platform page also in Word, Outlook, by email and as a mobile app
- Connections: Connector library via MCP, including iManage, SharePoint, Google Drive, Box, Slack, Teams and Outlook
- Setup: Access goes through sales: the home page offers a demo request. Administrators set up document repositories and roles.
- Pricing model: No public prices on the reviewed pages; the entry point is a demo request.

## Key features

### Agents for legal work

Agents handle work end to end, such as regulatory briefings or contract drafting from precedent, and return cited, review-ready output. You can preview the plan, adjust the scope and approve the work.

Source: https://www.harvey.ai/platform/agents

### Vault for document analysis

According to the page, a vault holds up to 100,000 documents. Review tables extract data points from many documents and compare them.

Source: https://www.harvey.ai/platform/vault

### Shared spaces

Spaces bring firm, clients and internal teams into one matter workspace. Guest accounts give outside users scoped access that administrators control.

Source: https://www.harvey.ai/platform/spaces

### Memory for preferences

Memory keeps working style and output preferences, not client or project details. You can review, edit and delete entries.

Source: https://www.harvey.ai/platform/memory

### Connector library

Tools such as iManage, SharePoint, Slack or Docusign can be connected via MCP according to the page. Administrators decide which connectors and which user groups are allowed.

Source: https://www.harvey.ai/platform/connector-library

### Command Center for administrators

Usage analytics by feature, practice group and region, plus SCIM profile sync and access management.

Source: https://www.harvey.ai/platform/command-center

### Task orchestration

In matter workspaces you assign tasks to Harvey agents, colleagues or clients and see where the work stands. Templates are intended for recurring matters.

Source: https://www.harvey.ai/platform/legal-work-orchestration

## Example workflows

### Prepare a regulatory briefing

- Your input: Internal policies and a question about a new regulation.
- Possible result: An agent drafts a cited memo for leadership.
- What to check: Review the plan up front and check citations and steps before the draft goes further.

### Run a matter in a shared space

- Your input: A space with template tasks for a recurring matter.
- Possible result: Tasks go to agents, colleagues or clients; finished work is reviewed in one place.
- What to check: Assign reviewers per task and check access for outside guests.

## Requirements and limits

### Access through sales

Requirement: The home page offers a demo request as the entry point and names neither prices nor a trial.

Source: https://www.harvey.ai

### Memory holds no client details

Limitation: Memory is personal and stores preferences and working style, not client or project details.

Source: https://www.harvey.ai/platform/memory

### Event-based triggers

Not yet confirmed: The agents page mentions schedules but describes no event-based triggers.

Source: https://www.harvey.ai/platform/agents

### Retention and deletion

Not yet confirmed: The security page mentions configurable retention policies and deletion but gives no default periods or deletion timelines.

Source: https://www.harvey.ai/security

### Privacy policy does not cover platform content

Not yet confirmed: By its own statement the privacy policy does not apply to inputs and outputs on the platform; it refers to the customer agreement for those.

Source: https://www.harvey.ai/legal/privacy-policy

## A good fit if …

- your firm or legal department wants a managed platform with agents, a vault and shared matter workspaces
- you want to review and approve agent work before it starts
- administrators should centrally control connectors and access by role

## Less suitable if …

- you want to compare prices or try it yourself without talking to sales
- you want to check data location and retention periods on public pages before a contract
- you need a self-run or open-source solution

Editorial assessment based on provider documentation. Not a product test.

## Usable AI models

Fixed choice across providers

Harvey works with models from OpenAI, Anthropic, Google and Mistral and also trains its own. You choose in the Model Selector or let Harvey decide with "auto". Some models are only open to workspaces that opt in; the pages checked do not mention your own keys or models.

- Model families named: OpenAI (GPT), Anthropic (Claude), Google (Gemini), Mistral, In-house model
- Model choice: Automatic or your choice
- Your own API key: Not stated
- Router or custom endpoint: Not stated
- Local models: Not stated
- Depends on: Admin approval

Documented · Sources accessed on: 2026-11-05

- [Harvey: Release notes, model updates](https://help.harvey.ai/release-notes/category/model-updates)

- [Harvey: Expanding Harvey's model offerings](https://www.harvey.ai/blog/expanding-harveys-model-offerings)

- [Harvey: Building a frontier in-house organization](https://www.harvey.ai/blog/building-a-frontier-in-house-organization)

## Europe & regional availability

Security page, privacy policy and Trust Center page reviewed. The Trust Center page named neither locations nor subprocessors when fetched.

Reviewed scope: Hosted platform; processing by customer requirement
Reviewed on: 2026-11-05

### Processing in the EU and Switzerland

With requirements

For customers with data-localisation requirements or preferences, Harvey offers processing in the EU and Switzerland or in Australia according to the security page. The cloud runs on Microsoft Azure. The page does not say which region applies by default.

Region covered: EU and Switzerland

Documented reason: Only for customers with such a requirement or preference.

Official source: https://www.harvey.ai/security

### Privacy policy and transfers

Availability still unclear

Harvey acts as a data processor for platform content. The policy does not apply to inputs and outputs on the platform. Personal data it covers may be transferred to servers in the US; it names Standard Contractual Clauses and the EU-US Data Privacy Framework among others.

Region covered: EU/EEA, United Kingdom and Switzerland

Documented reason: For customer content the policy refers to the customer agreement; location and transfer terms are set there, not on the reviewed page.

Official source: https://www.harvey.ai/legal/privacy-policy

### Trust Center

Availability still unclear

The Trust Center page lists SOC 2 Type 2, ISO/IEC 27001 and 42001, GDPR and EU AI Act among its entries. It does not explain what the GDPR and EU AI Act entries rest on.

Region covered: EU/EEA

Documented reason: The fetched page contains no statement on availability or data location in Europe.

Official source: https://trust.harvey.ai

## Accounts & permissions

Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.

Configuration covered: Hosted platform with administration features
Sources reviewed: 2026-11-05

### Account used

Documented: The security page lists SAML SSO among enterprise controls. Command Center syncs profiles via SCIM. Selected external collaborators get a guest account with scoped, admin-controlled permissions in Spaces, without needing full access to the Harvey platform.

Evidence: https://www.harvey.ai/security
Evidence: https://www.harvey.ai/platform/command-center
Evidence: https://www.harvey.ai/platform/spaces

### Admin approval

Documented: Administrators decide which connectors are enabled and which users or groups can use them. In Spaces, according to the page, nothing reaches a collaborator without explicit admin approval from both sides.

Evidence: https://www.harvey.ai/platform/connector-library
Evidence: https://www.harvey.ai/platform/spaces

### Data access & team sharing

Documented with conditions: Through connectors Harvey reaches systems such as iManage, SharePoint or Google Drive. Administrators can specify which tools need human confirmation. The reviewed pages do not describe exactly what rights agents hold inside those systems.

Evidence: https://www.harvey.ai/platform/connector-library

## Prices and included usage

Pricing checked: 2026-11-05

The home page names no prices and no free trial; the offered entry point is a demo request.

### Harvey

No public price; demo request only

Platform by arrangement with sales.

Price source: https://www.harvey.ai

Third-party pricing was not used as evidence.

Pricing details: https://www.harvey.ai

## Operation & oversight

Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.

Deployment covered: Hosted operation by the provider
Sources reviewed: 2026-11-05

### Files and working state

Documented with conditions: According to the page, a vault stores up to 100,000 documents. The security page mentions configurable retention policies; the reviewed pages name no default periods.

Evidence: https://www.harvey.ai/platform/vault
Evidence: https://www.harvey.ai/security

### Sign-ins

Documented: Sign-in via SAML SSO is listed among enterprise controls, and profiles can be synced via SCIM. The security page does not say which identity providers are supported.

Evidence: https://www.harvey.ai/security
Evidence: https://www.harvey.ai/platform/command-center

### Starting further runs

Documented with conditions: According to the page, agents can run reports, updates and reviews on a schedule you set. The agents page describes no event-based triggers. In matter workspaces you can assign tasks and tag Harvey in comments.

Evidence: https://www.harvey.ai/platform/agents
Evidence: https://www.harvey.ai/platform/legal-work-orchestration

### Human intervention

Documented: You can preview the plan, adjust the scope and approve work before Harvey begins. Harvey can bring you in when your judgment is needed, according to the page. For connectors, administrators specify which tools need confirmation.

Evidence: https://www.harvey.ai/platform/agents
Evidence: https://www.harvey.ai/platform/connector-library

### Detecting errors

Open in reviewed sources: According to the agents page, every step is logged for verification. The page does not describe how failures are handled.

Evidence: https://www.harvey.ai/platform/agents

### Resuming after failures

Open in reviewed sources: The reviewed agents page does not describe how interrupted runs are resumed.

Evidence: https://www.harvey.ai/platform/agents

### Editorial assessment

Harvey is a managed product with approvals before work begins and role control by administrators. For error handling, resumption and retention periods, the contract and Trust Center need to answer.

## Memory and conversation continuity

Scope: Memory feature for personal preferences
Sources accessed: 2026-11-05

Memory stores personal working style and can be reviewed, edited and deleted; the page does not describe storage location or retrieval.

### Storage and scope

Partly documented: Memory stores preferences and working style such as drafting conventions and output preferences, not client or project details. According to the page, memories like other workspace content are never used to train AI models. The page does not say where they are stored.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

### Selection and timing

Partly documented: You can set preferences upfront or as you work. The page does not describe the rules for what gets saved.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

### Inspect, edit and delete

Documented: Saved memories can be reviewed, edited or deleted anytime in one place. Administrators control how Memory is used across the organisation and can grant access by role.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

### Agent and channel boundaries

Partly documented: Memory is personal to each user. Saved preferences apply in the web app, Outlook, the Word add-in and playbook runs. The page describes no per-matter level.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

### Retrieval

Partly documented: Saved preferences are applied automatically according to the page. How they are retrieved or weighted is not stated.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

### Session reset and compaction

Not publicly established: Beyond deleting individual entries, the reviewed page says nothing about resetting or compaction.

- [Harvey: Memory](https://www.harvey.ai/platform/memory)

## Context

A contract with Harvey, legal documents and questions. The reviewed pages name no prices, retention periods or data-location details; they point to the customer agreement and the Trust Center.

Official website pages on platform, security and privacy reviewed. The Trust Center showed only parts when fetched. Feature descriptions come from the provider; example tasks are editorial. Hands-on tests are still pending.

## Open points

Europe, availability still unclear:
- Privacy policy and transfers
- Trust Center

Requirements and limits, not yet confirmed:
- Event-based triggers
- Retention and deletion
- Privacy policy does not cover platform content

## Sources

- https://www.harvey.ai

- https://www.harvey.ai/platform

- https://www.harvey.ai/platform/agents

- https://www.harvey.ai/platform/memory

- https://www.harvey.ai/platform/vault

- https://www.harvey.ai/platform/spaces

- https://www.harvey.ai/platform/command-center

- https://www.harvey.ai/platform/connector-library

- https://www.harvey.ai/platform/legal-work-orchestration

- https://www.harvey.ai/security

- https://www.harvey.ai/legal/privacy-policy

- https://trust.harvey.ai

## Provider

Harvey offers an AI platform for legal work.

- Operator: Harvey
- Organization: Provider
- Website: https://www.harvey.ai
- Reviewed: 2026-11-05

- Provider source: https://www.harvey.ai

- Provider source: https://www.harvey.ai/platform

[Sources and method](https://ki-agentenkompass.de/en/sources/)
