Everlaw MCP connectors let you securely search your Everlaw projects from the AI platforms you already use for analysis and drafting. For the full list of available MCP connectors, see the Connector reference section.
In this article, we'll cover:
- What you need before you connect
- How to set up an MCP connector
- Settings specific to individual MCP connectors
- What you can search, and how to get oriented in a project
- Background on how MCP connectors work
Requirements
To use an MCP connector, you need:
- An account with the AI platform for that MCP connector (see the Connector reference section for plan requirements)
- An Everlaw account with access to at least one project. Searches are limited to the projects and documents you can access in Everlaw.
Note
This connection is only available in our US environment. It is not available to Everlaw GovCloud users or those using our Canada, EU, UK, or Australia environments.
Data privacy and security
Your data privacy and security is our priority. Your use of Everlaw and our data privacy and security commitments are governed by our Terms of Service and Data Processing Addendum.
Your interactions within an AI platform are governed by your customer agreement with that platform's provider, which also outlines their security and data privacy practices.
Note
The connection is read-only. You can't change, export, or delete documents through an MCP connector.
Note
Anthropic (Claude) is contractually prohibited from using your Everlaw data to train its models.
Connector reference
MCP connectors are listed alphabetically. If an MCP connector has extra settings, the Connector-specific settings column points to them.
| MCP connector | Account required | Admin setup | Where it works | User connection docs | Connector specific settings |
|---|---|---|---|---|---|
| ChatGPT (OpenAI) | ChatGPT Enterprise workspace with the Everlaw plugin enabled for your role |
Required. A ChatGPT admin enables the Everlaw plugin and its connected app. See Plugin controls. |
Chat and Work (web, desktop, mobile); Codex in the desktop app; Codex CLI. Not supported in the Codex IDE extension. | Plugins | See Connector-specific settings > ChatGPT |
| Claude (Anthropic) | Claude account | Not required | Claude web and desktop apps | Use connectors to extend Claude's capabilities | See Connector-specific settings > Claude |
| Copilot (Microsoft) | Copilot account in an enterprise Microsoft tenant that has deployed the Everlaw connector |
Required. A Microsoft admin deploys the Everlaw connector. |
Microsoft 365 Copilot Chat, Copilot in Excel, Researcher agent | Federated connectors overview | None |
Set up an MCP connector
Step 1: Everlaw Organization Admin enables OAuth2
An Everlaw Organization Admin must enable OAuth2. To learn how, visit Organization Admin: Everlaw API.
Step 2: AI platform admin makes the MCP connector available (if required)
Some MCP connectors require an admin in the AI platform to make them available before users can connect. Check the Admin setup column in the Connector reference section. If it says Required, an admin in that AI platform must follow the linked instructions. If it says Not required, skip to Step 3.
Each user still signs in to Everlaw in Step 3, even if an admin set up the connector for them.
Step 3: Each user connects and signs in to Everlaw
- In your AI platform, find Everlaw in its list of connectors, plugins, or sources. See the user connection docs in the Connector reference section for where to look.
- Select the option to connect. You're redirected to the Everlaw login page.
- Sign in with your Everlaw credentials.
- When asked, select Allow to grant the AI platform read access to Everlaw.
You only need to connect once. You can now search your Everlaw projects from the AI platform.
Note
If you see an Access not enabled error, OAuth2 isn't enabled for your organization yet. Ask your Everlaw Organization Admin to complete Step 1.
For users on projects in multiple organizations
Search access depends on whether each project's owning organization has enabled OAuth2. If you can access projects in multiple organizations, you can only search projects in organizations that have enabled it.
When another organization enables OAuth2, its projects that you can access become searchable right away. You don't need to reconnect.
Connector-specific settings
Some MCP connectors have extra settings or tips. These sections are listed alphabetically.
ChatGPT
Choose how the plugin is offered to users
When a ChatGPT admin enables the Everlaw plugin, they can choose:
- Installed: Adds the plugin for everyone in the selected groups. Use this if you expect everyone to use Everlaw.
- Available: Lets users in the selected groups install the plugin themselves. Use this if only some users need it, so it doesn't clutter everyone's plugin list.
Claude
For Claude search tips and best practices, see Use Claude to Search and Understand Your Documents.
[Optional] Allow full-text access to documents
By default, Claude sees the search results Everlaw returns but not the full text of each document. For example, if you ask Claude to find documents that mention "birthday," Claude knows those documents contain "birthday" and can run more searches on them, but it can't read the rest of their text.
To let Claude read the full text of returned documents, a Claude Owner or Primary Owner (or a user with a custom role that grants access) must:
- Turn on Allow network egress in Claude's capability settings.
- Add esi-project.s3.amazonaws.com as an allowed domain.
Important
This setting applies to all users in your Claude organization.
What you can search
You can ask questions in natural language, and the AI platform translates them into Everlaw search syntax. You can search by:
- Document contents (text)
- Binders
- Bates/Control #
- Metadata terms, such as email participants, custodians, and dates
- Codes and freeform codes
- Productions
- Search Term Reports
- Extraction fields (field names only, not the extracted values)
Deep Dive
If your project is enrolled in Deep Dive, you can ask Deep Dive questions from your AI platform. For example: "On the Acme Review Project, use Deep Dive to research whether Beta Corp executives knew of the manufacturing defect before entering into the agreement."
Note
Deep Dive is not supported for Co-pilot (Microsoft).
Best practice: Get oriented in a project first
Before you search, ask the AI platform about the project's structure. This gives you useful context and helps it write more precise queries.
| Ask | The AI platform returns |
|---|---|
| What projects in Everlaw do I have access to? | Your projects. Start here, since other questions need a project. |
| What binders are in [project]? | Binders you can access |
| What processed uploads are in the project? | Processed uploads, useful for tracing documents to upload batches |
| What datasets are in this project? | Native datasets, so you can scope searches to specific collections |
| What metadata fields are in the project? | Searchable metadata fields and their formats |
| What Search Term Reports are in the project? | Search Term Reports |
| What codes are in the project? | Codes |
| What freeform codes are in the project? | Freeform code categories and their field types |
| What productions are in the project? | Productions |
| What extraction fields are in the project? | Extraction fields |
| What Bates prefixes are in the project? | Bates prefixes |
| What users are in the project? | Project users (Organization Admins and Project Admins only) |
For tool names and technical details, see Everlaw MCP Server.
Supplemental information: How MCP connectors work
You don't need this section to get started, but it explains how the connection works.
MCP connectors use the Model Context Protocol (MCP), a standard way to securely connect AI platforms to services like Everlaw. The Everlaw MCP server gives the AI platform a defined, read-only set of tools for searching your documents.
The MCP server authenticates with each user's own Everlaw permissions, so the AI platform never returns information you can't already access in Everlaw.
When you ask a question, the AI platform:
- Works out what information it needs from Everlaw.
- Translates your question into an Everlaw search and sends it through the MCP server.
- Uses the results to answer your question.
Because the AI platform understands natural language and can synthesize across documents, you can ask analytical questions, not just pull sets of documents.
The AI platform doesn't use the Everlaw API directly, but its tools mirror the Everlaw Search API endpoints. Developers can find implementation details in Everlaw MCP Server.