TL;DR: Wrike MCP can now suggest who should do the work. Ask Claude, ChatGPT, Copilot, Gemini, or any MCP client to staff a project, and Wrike returns an assignee recommendation for each task based on who's available and what the work requires. Nothing is assigned until you approve it. Available now in Wrike Labs for Pinnacle and Apex plans. New to Wrike MCP? Start with the Wrike MCP Server Overview.
Hi Wrike Community đź‘‹
Staffing a project is slow work, even with everything visible in Workload or Assignee Recommendations. Picking the right person for one task is straightforward. Picking them for a whole project isn't, because every assignment you make quietly removes an option from all the tasks you haven't staffed yet.
Today we're opening up intelligent resource allocation in Wrike MCP. Now you can just ask:
"Staff the Q4 Website Redesign project in Wrike with people from the Design team."
Wrike's own engine checks the project's tasks to see what effort, skills and other attributes are required, looks at who's available once existing commitments are counted, and comes back with a proposed assignee for every task plus the reasoning behind each one. Assignments are not made until you approve.
What it does
It adds one tool to the Wrike MCP server: recommend_resources, which shows up in your assistant as "Recommend assignees for Wrike items". You don't need to call it by name, just describe the staffing you want.
- Staffs a whole project, folder, or space, or just the specific tasks you name.
- Respects real capacity. Recommendations account for each person's availability and the work they're already committed to, and flag any period where a proposal would push someone into overload.
- Matches skills to work. Where your tasks carry requirements (job roles, skills, and other custom fields in user profiles), candidates are matched against them. You can ask for strict matching, or tell your assistant to relax the requirement when the perfect person isn't free.
- Narrows the candidate pool. "Use only the design team," "only people in this space," "use assets", all work in plain language.
- Shows the workload impact. Ask for the details and you'll see capacity, what's already committed, what's being proposed, and the resulting projected load, by day or by week.
Nothing changes until you approve. Your assistant proposes; nothing changes in Wrike until you explicitly confirm. Then it applies the assignments for you in one go.
Prompts to try
- "Staff [project] in Wrike and show me who you'd put on each task."
- "Recommend assignees for the tasks in [project], but only from the Creative team."
- "Give me two staffing options for [project]: one where everyone matches all required skills, and one that keeps everyone under capacity, even if up to 20% of requirements aren't met."
- "Who could take [task] without going over capacity?"
- "Can [person] take this [task]?"
- After reviewing the options: "Apply the second option." (This assigns people in Wrike.)
- "Why is [person] recommended here?"
Get the same staffing view every time (optional)
You can already ask your assistant to show recommendations as a visual chart or a Klaxoon Whiteboard. Two skills make that view consistent every time:
Before you start
- Plan: Pinnacle or Apex.
- Wrike Labs: enable for your individual profile in Wrike Labs - “Find the best assignees with intelligent resource allocation”.
- Wrike MCP connected to your AI assistant. If you haven't set that up yet, pick your assistant in the Wrike MCP Server Overview and follow the setup guide. It takes a couple of minutes with OAuth.
- MCP v2. Resource allocation is a v2 tool. If you configured a custom connector yourself, make sure your Server URL is
https://mcp.wrike.com/v2. - Refresh your connector. MCP clients cache the tool list from the first time they connect, so a new tool may not appear until you refresh, usually by disconnecting and reconnecting the Wrike connector, or manually selecting a refresh tool option. See the refresh table in the overview for your specific client.
- Your data: recommendations are only as good as what's in the workspace. Tasks need effort estimates and dates, and people need their work schedules set up. Matching on job roles, skills and other requirements only applies to tasks assigned to job roles or where task requirements are filled in.
Once that's in place, ask your assistant to staff your next project.
What it can't do yet
- About 500 task assignments per request. MCP puts a ceiling on how much data a tool can send back in one response, so a single request currently returns up to roughly 500 assignments. For a bigger backlog, staff it in pieces: one subproject or one phase at a time.
- Up to 1,000 people in the candidate pool. The tool can weigh a maximum of 1,000 users per request. If your account or the group you point it at is larger than that, add a filter so fewer people go into the pool, such as a specific team or a space.
Tips for better results
- Name the scope. "Staff project X" works better than "staff my projects." If the scope is too broad, the tool will ask you to narrow it: a smaller project, specific tasks, or a single team.
- Name the people, too. Say who should be considered, for example "only the Creative team" or "people in the Marketing space." You'll get more relevant recommendations, and on a large account it keeps you inside the 1,000-user limit.
- Say what matters most. Capacity or skills and job roles: tell your assistant what you're optimizing for and it will weigh the request accordingly. Other dimensions are coming soon!
- Review before applying. Treat the recommendations as a draft you edit. Check the overload and missing skills flags before you confirm.
đź’¬ We'd love your feedback
This is a Labs feature, which means it's still taking shape and your input changes it. Tell us what you're staffing, where the recommendations landed well, and where they missed. We're reading every reply.
Prefer to talk it through live? Book a call with the team and show us how you're using it.
Related
