All articles

Wrike Model Context Protocol (MCP)

TL;DR

Wrike MCP lets AI assistants like Copilot, ChatGPT, and Claude securely access and update your Wrike workspace, so you can search, manage, and create tasks and projects in natural language without opening Wrike.

Table 21. Availability

Free

Team

Business

Pinnacle

Apex

Availability: Free, Team, Business, Pinnacle, Apex. ; Unavailability: ;

Overview

Wrike offers a Model Context Protocol (MCP) server that lets AI assistants read and update your Wrike data.

You can query projects, manage tasks, and navigate folders from MCP‑compatible tools such as Microsoft Copilot, Claude, ChatGPT, Cursor, and others.

The Model Context Protocol (MCP) is an open standard that lets AI assistants interact with external tools and data sources, including your Wrike workspace.

Why Use Wrike MCP

Wrike MCP helps you work with Wrike through your AI assistant instead of switching apps.

You can use it to:

  • Boost productivity

    • Manage tasks, projects, and folders from your AI assistant without opening Wrike

  • Set up quickly

    • Connect through simple OAuth with one-click sign‑in for supported tools (Microsoft Copilot, ChatGPT, Claude, and more)

  • Work directly with your data

    • Let AI assistants read and write Wrike data without exports or copy‑pasting

  • Use AI‑ready data

    • Take advantage of data that’s formatted specifically for AI agents

What You Can Do With Wrike MCP

Once you connect Wrike MCP in a supported AI tool, you can:

  • Report and act on status

    • Ask for status summaries of team tasks

    • Update statuses, reassign work, or change dates directly from your AI assistant

  • Prioritize work

    • Get a list of urgent or high‑impact tasks based on importance and due dates

  • Search and find answers

    • Use natural language to find information across Wrike’s folder and project hierarchy

  • Create work structures

    • Create folders, projects, and tasks through conversation

  • Convert meeting notes to tasks

    • Turn action items into Wrike tasks with assignees and due dates

  • Onboard into your workspace

    • Explore spaces, projects, custom fields, and workflows conversationally

and more, depending on what your AI tool supports.

For detailed steps to connect your AI assistants to Wrike, see the Wrike MCP server documentation.

FAQ

For IT Security

Does Wrike MCP store or cache any customer data?

No. Wrike MCP is fully stateless. It acts as a real-time pass-through to the Wrike API. No data is cached, logged, or stored by the MCP layer. All data persistence happens in Wrike's existing infrastructure.

What authentication method does MCP use?

OAuth 2.0 - the same standard Wrike uses for all API integrations. No static API tokens. Users authenticate through Wrike's OAuth flow and tokens are scoped to their individual permissions. Tokens can be revoked at any time.

Is all communication encrypted?

Yes. All communication uses HTTPS with TLS 1.2+. The remote SSE endpoint is served from wrike.com/app/mcp/sse behind Wrike's existing TLS infrastructure.

Can MCP access data beyond a user's permissions?

No. MCP enforces Wrike's full RBAC model. A user connecting via MCP can only access exactly what they can access in the Wrike web UI. There is no privilege escalation. Admin-restricted content stays restricted.

Does MCP require Wrike's AI Addendum?

No. The Wrike AI Addendum covers Wrike's in-app AI features (e.g., Agent Builder, Wrike Copilot Wrike Assist), which use Azure OpenAI. MCP is a data access protocol - no AI models run on Wrike infrastructure. MCP is architecturally equivalent to any other Wrike API integration.

Does Wrike see our AI prompts or conversation history?

No. Wrike MCP only receives structured API calls (e.g., "get tasks filtered by status=Active"). It has no visibility into the AI conversation, prompts, or reasoning that generated those API calls.

Is there audit logging for MCP actions?

Yes. All actions performed via MCP appear in Wrike's standard activity stream under the authenticated user's account. They are indistinguishable from actions taken through the web UI - same user attribution, same timestamps, same change records.

Can we restrict or revoke MCP access?

Yes. Admins can revoke OAuth tokens at any time. When a user is deprovisioned from Wrike, their MCP access is automatically revoked. OAuth tokens can also be managed centrally through a registered OAuth app.

Is SSO compatible with MCP?

Yes. MCP uses Wrike's standard OAuth 2.0 flow, which respects your SSO configuration. If your organization uses SAML/SSO for Wrike login, the OAuth flow will go through the same identity provider.

What is network architecture? Can we IP-allowlist?

MCP uses a cloud-hosted remote SSE server at wrike.com/app/mcp/sse (EU: app-eu.wrike.com/app/mcp/sse). Since it runs on Wrike's existing infrastructure, your existing Wrike firewall rules and IP allowlists apply. No additional network configuration is needed.

What is the token rotation policy?

OAuth 2.0 tokens are managed through Wrike's standard infrastructure. Admins can revoke tokens at any time. We recommend a 90-day rotation policy for security best practices.

Can we start with read-only access first?

Yes. Read tools execute freely with no risk. Admins can evaluate read-only usage before enabling write tools. This is a recommended approach for pilot deployments.

What happens if data is accidentally modified?

No delete operations exist in Wrike MCP. Create actions are additive (non-destructive). Update actions are idempotent (safe to retry). All changes are tracked in the audit trail under the user's account.

For Legal

Do we need a new DPA for MCP?

No. MCP uses Wrike's existing API infrastructure. Your existing Wrike DPA covers MCP data access. You may need a separate DPA with your AI provider (e.g., Anthropic, Microsoft) for AI processing, but that is independent of Wrike.

Does Wrike train AI models on our data?

No. Wrike does not use customer data for AI model training. MCP doesn't involve any Wrike AI models. Whether your AI provider uses data for training depends on your agreement with that provider (e.g., Claude Enterprise does not train on customer data).

What are the GDPR data transfer implications?

MCP does not create new data transfer flows with Wrike. Data accessed via MCP follows the same paths as the Wrike API. EU customers can use the EU endpoint (app-eu.wrike.com/app/mcp/sse). Data transfer to your AI provider is governed by your agreement with that provider.

Which Wrike certifications apply to MCP?

All of Wrike's certifications apply because MCP runs on Wrike's existing infrastructure: SOC 2 Type II, ISO 27001, GDPR, and CSA STAR. MCP is not a separate product - it's an additional API access layer within Wrike's certified environment.

Do we need Wrike's AI Addendum?

No. The Wrike AI Addendum is for Wrike's in-app AI features (e.g., Agent Builder, Wrike Copilot ), which use Azure OpenAI within Wrike. MCP involves no AI processing on Wrike's side. It's a data access protocol in the same category as the Wrike REST API.

Top

Related articles