Wrike MCP Server Overview
Wrike MCP v2 is now in BetaWe're rolling out MCP v2: faster, leaner, and smarter about your data (unified search, cleaner context, lighter navigation that cuts token use by ~25%, bulk actions, inbox, writing comments and more). We have already updated the whole documentation just pointing out to the MCP v2 throughout.
How can you test it today? If you are using a custom OAuth connection (where you configured Server URL, OAuth Client ID and Client Secret when adding an MCP to your AI Assistant), just follow the setup guides below, we have updated the Server URL there. Remember that you will be configuring a separate custom connector, so you will need to remove the MCP v1 connector once you make the MCP v2 available. On most AI assistant plans, only the account administrator for ChatGPT, Claude, Microsoft Copilot Studio, etc. can add or update a custom OAuth connector, so if that's not you, please ask your administrator to set it up.
Instead, if you are using the built-in Wrike connector in Claude, ChatGPT, or Gemini Enterprise, you will not be able to test it until it is generally available. We're submitting v2 to the AI marketplaces shortly. Once it's approved, the pre-configured Wrike connector updates to v2 automatically, so you will not need to reinstall anything.
Which tools are available? See Available tools on Wrike MCP
Which use cases it unblocks? For inspiration, see Typical Use Cases of Wrike MCP
Wrike offers a Model Context Protocol (MCP) server, accessible via Apps & Integrations, that enables AI assistants to access and interact with your Wrike workspace data— query projects, manage tasks, and navigate folders. It works with Microsoft Copilot, Claude, ChatGPT, Cursor, and any other MCP-compatible clients.
The Model Context Protocol (MCP) is an open standard that lets AI assistants interact with your Wrike workspace.
Why use Wrike MCP?
Productivity boost
Manage Wrike tasks without switching contexts from your AI assistant
Easy setup
Connect through simple OAuth, with one-click sign-in, for supported AI tools including Microsoft Copilot, ChatGPT, Claude, and others
Access to your data
AI assistants can read and write your Wrike data without manual exports or copy-pasting
Optimized for AI
Built specifically for AI agents with efficient data formatting
What can you do with Wrike MCP?
- Report & act on status — Get quick status updates on team tasks and act on them without opening Wrike
- Prioritize work — Get an overview of what's urgent based on importance and due dates
- Search & find answers — Find specific information across Wrike's folder hierarchy using natural language
- Create work structures — Set up folders, projects, and tasks through conversation
- Convert meeting notes to tasks — Turn action items into tracked Wrike tasks with assignments and due dates
- Workspace onboarding — Explore spaces, projects, custom fields, and workflows conversationally
- and much more
To see a tools that are supported on Wrike MCP, check Available tools on Wrike MCP.
Detailed setup instructions for each AI assistant
Alternative authentication: Permanent Access Token
OAuth is the standard authentication method that we recommend for the majority of the cases where you connect your AI Assistant with Wrike MCP server. This will avoid users having to manage tokens or mcp configurations, and the only thing that the user will need to do is to log in inside wrike when the authentication is triggered.
For fully automated workflows, cloud-based coding agents that run without human interaction or individual use, the Permanent Access token can be used.
Each user generates an App and then generates a Permanent Access Token that inherits their Wrike permissions. See Alternative Authentication Setup for step-by-step instructions.
Usage examples
Example 1: Report status
Prompt: "What tasks are assigned to Carol that are overdue?"
Result: Returns overdue tasks with title, status, due date, and links. Follow up with "Mark the first task as completed" to take action directly.
Example 2: Create work structure
Prompt: "Create a folder called 'Client: Acme Corp' in the Project Team space. Create 3 projects inside: Onboarding, Implementation, Support. For Onboarding, add tasks: Kickoff Call, Requirements Gathering, Contract Signing. Assign all to Katerina, due within 2 weeks."
Result: Complete folder/project/task structure created with assignments and due dates.
Example 3: Convert meeting notes to tasks
Prompt: "Action items from today's meeting: 1) Update Campaign brief — Susan; 2) Arrange photoshoot — Jonathan; 3) Meeting with models — Theresa. Create tasks in Marketing space, Backlog folder, assign appropriately, all due next Friday."
Result: Three tasks created with correct assignees and due dates.
Important considerations
- LLMs are non-deterministic — the same query may produce slightly different responses each time. This is normal AI behavior, not a bug in the integration.
- Be specific in your queries — include permalinks, project names, folder locations, or task details when known for best results.
- If the first response isn't perfect, refine your query with more context.
Troubleshooting
Problems with the default pre-configured Wrike MCP available in Claude, ChatGPT, and Gemini
The connector is available or published by an administrator, but is not visible to users
- Make sure you have checked all available connectors. Sometimes the list is shuffled or reordered alphabetically after a connector is connected.
- Test with more than one user to confirm that the issue is not specific to a single user.
Some tools are not available in the MCP
- Check whether the tool you are interested in is listed here: MCP List of tools
- Check with the AI administrator to confirm that all required tools are enabled for the account on the AI assistant administration side.
Problems with a custom Wrike MCP that you set up manually using OAuth configuration data
I do not have access to “Apps & Integrations” or cannot create a new app
- Make sure you have the required API permissions in Wrike. The Account Owner always has these permissions.
I cannot set up a custom connector
- Check whether “Developer Mode” needs to be enabled to connect a custom connector.
- You may need support from the AI administrator if you do not have sufficient permissions to set up a custom MCP.
Authentication error with OAuth
- Make sure that the browser opened during authentication is the one where you are logged in to Wrike.
- Make sure that the redirect URLs are configured correctly for the OAuth client.
- If you experience re-authentication issues, try switching the MCP server URL from streamable HTTP (https://mcp.wrike.com/app/mcp/stream) to SSE by changing the Server URL to: https://mcp.wrike.com/app/mcp/sse.
- Contact your AI assistant administrator so they can check whether the connector’s OAuth configuration is set up correctly.
Authentication error with Permanent Access Token
- Verify that the Permanent Access Token is set correctly in the AI assistant configuration.
- Generate a new token if the current one is not working.
- Generate a new App if the current one is not working, and then generate a new token.
Inconsistent responses
- Remember that LLMs are non-deterministic by nature.
- Add more specific context to your queries.
- Use system prompts to guide the AI assistant’s understanding of Wrike.
Some tools are missing, or new tools don't appear after an update
The Wrike MCP server exposes a fixed set of tools — you don't add or remove individual tools per connection. Your AI client caches that tool list the first time it connects, so after Wrike ships new tools (for example, the v1 → v2 update), your client may keep showing the old list until you refresh it. This is standard MCP client behavior, not specific to Wrike.
To load the latest tools, refresh the Wrike connector in your client:
| AI client | How to refresh |
|---|---|
Claude / Claude Code | Disconnect and reconnect the Wrike connector (Settings → Connectors). In Claude Desktop, restart the app. An older connection may appear as a "Custom" connector until you re-add it. |
ChatGPT (personal / custom) | Settings → Connectors → your Wrike connector → Refresh. |
ChatGPT (Business / Enterprise) | An admin refreshes it: Workspace Settings → Apps → the Wrike app → Action control → Refresh. New tools are disabled by default and shown as a diff — the admin must enable them. Until then, a changed tool can error. |
Microsoft Copilot Studio | On the agent's Tools tab, make sure Allow all is on (when it's off, newly added tools are off by default — enable them individually). If the old list persists, remove the Wrike MCP tool and add it again (Add a tool → Model Context Protocol), then Publish the agent. |
Gemini Enterprise | An administrator reconnects the Wrike connector for the organization. |
Cursor / VS Code | Reload the window, or restart the editor. |
| Other MCP clients | Check the client's documentation for how to refresh the MCP tool list (usually disconnect/reconnect or restart). |
Privacy policy and terms of use
Your privacy matters. See the Wrike Privacy Policy and Wrike Terms of Use for details on how your data is handled.
- API tokens and OAuth tokens are stored locally on your device
- All communication is encrypted via HTTPS
- The MCP server only accesses data you have permissions for in Wrike
- No data is stored or cached by the MCP server
Support
- Help Center: Contact Wrike Support (live chat, request form, phone for Premium plans)
- Community: Submit feature requests through Wrike Community
Updated about 14 hours ago