Client Onboarding with MCP: A Step-by-Step Playbook

Client onboarding is the run of setup work after a contract is signed: a new project, a kickoff task list, owners, due dates and a welcome note. An account manager or project manager usually does it by copying last month's project and editing it by hand. With MCP (Model Context Protocol), an AI app such as Claude or ChatGPT can read your onboarding checklist and create the project and tasks in your work tool, while you approve each change.

This playbook uses three official, vendor-hosted servers: Asana, monday.com and Linear. Pick the one your team already uses. Everything below was checked against each vendor's docs on 28 October 2026. eSEOspace did not test these servers.

What you need

You need one MCP client and one project tool server.

Notes from the vendors' pages:

• Asana. Asana's MCP docs say the old beta endpoint at https://mcp.asana.com/sse is deprecated, with a listed shutdown date of 05/11/2026. Use the V2 URL. Our Asana MCP server profile has the details.
• monday.com. The Platform MCP tools page lists 64 tools across 15 categories. See the monday.com MCP server profile.
• Linear. Linear's MCP docs say the tools are for "finding, creating, and updating objects in Linear like issues, projects, and comments." See the Linear MCP server profile.

For the client, Claude, ChatGPT, Cursor and other apps work with all three. Each vendor's docs give the setup steps for each client.

The workflow

1. Write the onboarding checklist once. Keep it as a plain document: phases, tasks, default owners by role, and due-date offsets from the kickoff date. This is the source the AI copies from. It stays under your control.
2. Read before you write. Ask the AI to find the right workspace and any existing project for the client. On monday.com that is search or list_workspaces. On Asana, use a read request like Asana's own example, "List all sections in the Product Launch project." On Linear, connect to /mcp/readonly for this step.
3. Create the container. On monday.com, create_board creates a board and create_group creates groups for each phase. On Asana and Linear, ask for a new project and check which tool your client calls.
4. Add the tasks. On monday.com, create_item creates items or subitems and can duplicate an existing item. Asana's docs give the example "Create a new task in the Marketing project assigned to me." Linear's docs describe turning a plan into projects, issues and sub-issues.
5. Assign owners and dates. On monday.com, list_users_and_teams finds people and change_item_column_values sets owner and date columns. In Asana and Linear, the AI updates the tasks it just created.
6. Attach the welcome pack. monday.com's create_doc creates a WorkDoc in a workspace or on an item. Use it for the kickoff agenda or access checklist.
7. Post the handoff note. On monday.com, create_update posts a comment on an item. Keep this step for last, after a person has checked the board.
8. Review and fix. Read the new project back and compare it with the checklist. Ask the AI to list anything missing rather than to "fix everything."

Example prompts

• "Read our onboarding checklist below. Then find the Clients workspace in monday.com and tell me whether a board for Acme Dental already exists. Do not create anything yet."
• "Create a board called Acme Dental Onboarding with groups for Kickoff, Access, Build and Launch. Show me the plan before each create call."
• "Add the Kickoff tasks from the checklist as items in the Kickoff group. Set due dates from a kickoff date of 3 November, and leave owners blank."
• "In Asana, create a new project for Acme Dental from these sections, then list all sections so I can check them."
• "In Linear, turn this onboarding plan into a project with one issue per step. List the issues you would create first."

Guardrails

• Start read-only. Use Linear's /mcp/readonly endpoint or a Read-only API key while you learn the flow. For the others, run the read steps first and approve writes one by one.
• Keep approvals on. The MCP tools specification says there should always be a human in the loop who can deny tool calls. Onboarding is a good place to keep that rule strict, because tasks notify real people.
• Use a scoped account. monday.com says every call runs as the signed-in user and suggests limiting MCP use to specific workspaces or boards, per its security overview. The same idea applies to Asana and Linear.
• Know the admin controls. Asana says Enterprise+ and Legacy Enterprise customers can allow or block MCP clients through app management. monday.com admins can approve dynamic connectors before tokens are issued.
• Watch revocation timing. monday.com says that during early access, a revoked dynamic connector can keep a valid token for up to 24 hours.
• Keep the judgment human. Pricing, scope, contract dates and anything said to the client stay with your team.

The MCP security hub and our lesson on least privilege in practice cover these controls in more depth.

What to measure

Track a baseline for a few onboardings done by hand, then compare:

• Time from signed contract to a ready project.
• Checklist steps missed or added late.
• Edits a person had to make after the AI created the project.
• Writes the reviewer rejected, and why.
• For monday.com, API calls used, since MCP calls count toward your daily API limit.

If rework stays high, improve the checklist before you change the prompts.

Next step

More task guides like this one live in the MCP playbooks hub. If you want help picking a server and setting permissions for your team, start with an MCP readiness audit.

Frequently asked questions