iManage Work

Name: iManage Work MCP Server
Availability: InStock
Author: iManage Work

Productivity14 tools

Search, browse and read a law firm's document management system from your AI assistant. Fourteen read tools in Anthropic's directory — and iManage's own live OAuth descriptor lists twenty-three, including seven writes the listing omits. iManage states ethical walls and permissions are enforced.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on 19 August 2026.

Anthropic states this reflects the level of review a connector received, not a security audit.

Connect iManage Work via MCP

https://cloudimanage.com/mcp/work

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

iManage Work Tools & Capabilities (14)

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

download_document returns document text inline, with no documented size cap. iManage documents system-created chunking for large documents, so length is not a barrier — a long privileged document arrives in full, across responses, in the model's context.
Anthropic's directory lists 14 tools, all reads; iManage's live OAuth descriptor lists 23, seven of them writes. The seven writes appear in neither the directory nor iManage's published tool reference. Read the descriptor, not the listing, to know the designed surface.
The "Read and write" permissions label does not describe what a user can authorise today. iManage states only read permissions are currently available and selected by default, and refers to write tools arriving in upcoming releases.
Ethical-wall and permission enforcement is documented by iManage, not verified by us. We did not authenticate and called no tool. A firm should test its own walls against its own tenant.
Permission inheritance is a ceiling, not a floor. An agent connected as a user with broad cross-matter access has exactly that access.
Only one OAuth scope is requestable. The four finer permissions iManage publishes sit under a non-standard imanage_scopes key, absent from scopes_supported. A client cannot narrow the grant by asking for less; narrowing happens in Control Center.
Cloud only. iManage states the MCP server is unavailable for iManage Work at imanage.work.
Administrators cannot view other users' MCP sessions. iManage documents this as deliberate. The document history timeline, not the session page, is the compliance surface.
Only two calls are documented as writing a timeline entry — download_document on this connector and the Search connector's fetch. get_rows_from_csv_document reaches whole spreadsheets and is not among them; we could not establish whether another mechanism audits it.
iManage disclaims control over model output, stating response quality and behaviour depend on the model provider and version, and directing customers to ask their AI vendor what data leaves iManage.
Pagination caps results at 500 across search, get_recent_documents, get_recent_workspaces and others, at 100 per page by default; iManage states search does not support a custom page limit. get_container_children returns at most 500 children per call.
No prompts are listed in Anthropic's snapshot. We verified this against the record directly, because a sibling page this run found prompts where a brief said there were none.
We did not read live tool schemas. The endpoint returned 401 to an anonymous handshake, so tool names come from Anthropic's snapshot, iManage's tool reference and iManage's OAuth descriptor —

Frequently asked questions

Does the iManage MCP server respect ethical walls and document permissions?

iManage states it does. Its documentation says existing security policies, permissions and audit trails are maintained, and that a user without access to a document in iManage will not have access through the MCP server. Its launch announcement names ethical walls specifically as respected. That is a vendor claim about its own enforcement, not something we could test without authenticating.

What does download_document return from iManage Work?

The text content of a document, inline in the response. iManage documents the tool as downloading text content by document ID, and describes system-created chunking when a document is too large to return in one response. So privileged document text lands in the model's context rather than behind a link. iManage records each call on the document's history timeline.

Can the iManage MCP connector write to or modify documents?

Not through Claude today, though write tools exist. iManage's setup instructions state that only read permissions are currently available and selected by default. Its live OAuth descriptor lists seven write tools, and its administrator FAQ refers to write tools arriving in upcoming releases. Administrators can already choose Read, Read and Write, or No access per service.

How many tools does the iManage Work MCP server have?

It depends which source you read, and they disagree. Anthropic's directory snapshot lists fourteen, all reads. iManage's published tool reference documents the same fourteen. But iManage's own live OAuth protected-resource descriptor enumerates twenty-three tools with explicit read or write types — sixteen reads and seven writes. The seven writes appear in neither published list.

Can on-premises iManage customers use the MCP server?

No. iManage states the MCP server is supported only for iManage Work at cloudimanage.com, and explicitly unavailable for iManage Work at imanage.work. Firms running iManage outside that cloud service cannot connect this way. iManage also states the server uses existing iManage Cloud infrastructure and requires no additional on-premises hardware.

Can a firm control which AI clients reach its iManage documents?

Yes, through iManage Control Center. iManage documents an MCP Settings page where administrators allow-list which MCP-enabled clients may connect, with unauthorized clients flagged for review. Administrators also build access policies setting Read, Read and Write, or No access per iManage service, optionally per client, and restrict which users and groups can use the server at all.

Does iManage document what happens to document content sent to an AI model?

Partly, and it is candid about the boundary. iManage states data remains in iManage and that the server is a secure bridge, and that revoking access destroys all tokens with no client data retained. But it also tells customers to work with their AI vendors on what data leaves iManage, and states it has no control over model responses.

Sources

iManage Work connector tool reference — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Tools_for_iManage_Work_connector.html (fetched 2026-08-19; the documentation URL in Anthropic's directory. Documents exactly 14 tools in four groups, with a description and sample prompt for each. Source for the download_document text-and-chunking behaviour, the fetch document-or-workspace semantics, the search query-construction guidance, the 1,000-version cap on get_document_versions and the 500-child cap on get_container_children) · retrieved 2026-08-19
iManage MCP Server Help, introduction — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Introduction_to_iManage_MCP_Server.html (fetched 2026-08-19; source for the permission, audit-trail and data-remains-in-iManage statements) · retrieved 2026-08-19
iManage MCP Server Help, understanding the integration — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Understanding_the_integration.html (fetched 2026-08-19; the three-step token-introspection and access-control flow) · retrieved 2026-08-19
iManage MCP Server Help, adding connectors for Claude — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Adding_MCP_connectors_for_Claude.html (fetched 2026-08-19; states twice that only read permissions are currently available and selected by default; source for the custom-subdomain URL form and the sibling connector URLs) · retrieved 2026-08-19
iManage MCP Server Help, tools available for each connector — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Tools_available_for_each_MCP_connector.html (fetched 2026-08-19; the four connectors and the pagination caps of 500 total and 100 per page) · retrieved 2026-08-19
iManage MCP Server Help, sessions — https://docs.imanage.com/ai/mcpserver-user-help/en-US/Viewing_your_session_details_and_deleting_your_sessions.html (fetched 2026-08-19; the session console at cloudimanage.com/mcp/web/, its recorded fields and deletion) · retrieved 2026-08-19
iManage MCP Server Administration Help, FAQ — https://docs.imanage.com/ai/mcpserver-admin-help/en-US/iManage_MCP_Server_FAQ.html (fetched 2026-08-19; source for the sensitive-documents answer, the timeline-entry table naming download_document, the note about write tools in upcoming releases, token destruction on revocation, the disclaimer of control over model output, and RFC 7591 dynamic client registration with HTTPS-only redirects) · retrieved 2026-08-19
iManage MCP Server Administration Help, requirements — https://docs.imanage.com/ai/mcpserver-admin-help/en-US/Requirements.html (fetched 2026-08-19; states support for iManage Work at cloudimanage.com and unavailability at imanage.work) · retrieved 2026-08-19
iManage MCP Server Administration Help, configuring MCP settings — https://docs.imanage.com/ai/mcpserver-admin-help/en-US/Configuring_MCP_settings.html (fetched 2026-08-19; the disclaimer gate, client allow-listing with flagging of unauthorised clients, and access policies) · retrieved 2026-08-19
iManage MCP Server Administration Help, access policies — https://docs.imanage.com/ai/mcpserver-admin-help/en-US/Managing_and_adding_MCP_access_policies.html (fetched 2026-08-19; Read / Read & Write / No access per service, default Read, per-client permission sets, tool-level inspection) · retrieved 2026-08-19
iManage MCP Server Administration Help, adding the MCP application — https://docs.imanage.com/ai/mcpserver-admin-help/en-US/Adding_the_iManage_MCP_application_in_iManage_Control_Center.html (fetched 2026-08-19; the two application registrations and their independent access lists) · retrieved 2026-08-19
Live RFC 9728 protected-resource descriptor — https://cloudimanage.com/.well-known/oauth-protected-resource/mcp/work (fetched 2026-08-19; the form named by www-authenticate. HTTP 200, 3,786 bytes. Resource https://cloudimanage.com, one entry in scopes_supported, an imanage_scopes object of four permissions, and a 23-entry tool manifest typed read or write. Root and path-insert forms return HTTP 404 with 15-byte and 9-byte plain-text bodies) · retrieved 2026-08-19
Live RFC 8414 authorization-server metadata — https://cloudimanage.com/.well-known/oauth-authorization-server (fetched 2026-08-19; issuer https://cloudimanage.com, S256 PKCE, grant types authorization_code and refresh_token, a registration_endpoint, revocation_endpoint and userinfo_endpoint, and four service scopes. Both path variants return 404) · retrieved 2026-08-19
Live anonymous probe — 2026-08-19: an anonymous initialize POST to https://cloudimanage.com/mcp/work returned HTTP 401 with header x-error-code: IM_E_MISSING_TOKEN and www-authenticate: Bearer resource_metadata="https://cloudimanage.com/.well-known/oauth-protected-resource/mcp/work", scope="imanage_work_mcp_scopes". No tool was ever called and no authentication was attempted · retrieved 2026-08-19
iManage launch announcement — https://imanage.com/resources/resource-center/news/mcp-server-available-broader-ai-ecosystem/ (fetched 2026-08-19; dated 14 May 2026. Source for the ethical-walls statement, the permission-bound and fully-logged claim, and the instruction that customers work with AI vendors on what data leaves iManage) · retrieved 2026-08-19
docs.imanage.com robots.txt — https://docs.imanage.com/robots.txt (fetched 2026-08-19; User-agent: * with disallows on an admin path and many superseded product-help version directories. No Content-Signal directives and none in the response headers. docs.imanage.com/llms.txt returns HTTP 404) · retrieved 2026-08-19
imanage.com robots.txt — https://imanage.com/robots.txt (fetched 2026-08-19; 93 bytes, two disallows and one sitemap declaration. No Content-Signal axes. imanage.com/llms.txt returns a zero-byte HTTP 404) · retrieved 2026-08-19
Anthropic connector directory — https://claude.ai/directory/bfc6b32d-8cdb-451d-8132-d8a16bfe9f9c (snapshot 2026-08-16; 14 tool names, empty prompt_names verified directly on the record, permissions "Read and write", author iManage LLC, partner tier, added 2026-05-11) · retrieved 2026-08-16