A paid plan
The MCP server is part of the Professional, Team and Enterprise plans. Workspaces on Starter cannot connect.
The Meridian MCP server lets an AI tool you already use, such as Claude, ChatGPT or Cursor, read your project data and, if your admin allows it, log time and save status reports. It works as you, with your permissions.
Available on Professional, Team and Enterprise plans.
https://www.meridian4pm.com/api/mcpAlways use the www address. Without www, the address redirects and AI tools do not follow the redirect.
What is MCP
The Model Context Protocol (MCP) is an open standard that lets an AI tool call a list of tools exposed by another service. Meridian publishes one address. Your AI tool connects to it, you approve what it may do, and it can then answer questions such as "which milestones are overdue?" from your live data.
Meridian makes no AI calls of its own for this. Your AI tool asks, Meridian answers, and nothing is sent anywhere else.
Before you connect
The MCP server is part of the Professional, Team and Enterprise plans. Workspaces on Starter cannot connect.
An admin switches MCP on under Settings, Integrations. Writes are a separate switch, off until an admin allows them.
Your AI tool acts as you. Viewers can connect to read but never write. Guests cannot connect.
Connect your tool
Vendor menus change. If a screen does not match, check the vendor's current documentation; the server address is the same everywhere. When you sign in, Meridian shows which app is asking, what it can read, and, if your admin allowed writes, optional boxes for logging time and saving status reports that you can untick.
Add Meridian as a custom connector. You sign in to Meridian in your browser; no token is copied.
Custom connectors depend on your Claude plan and may need a workspace owner to switch them on. Check Anthropic’s current documentation, because menu names change.
One command adds the server. Then sign in once from inside Claude Code.
claude mcp add --transport http meridian https://www.meridian4pm.com/api/mcpAdd Meridian as a custom app that signs in with OAuth.
Custom apps depend on your ChatGPT plan and workspace settings, and the menus change often. Check OpenAI’s current documentation.
Add Meridian to your MCP settings as a remote server.
{
"mcpServers": {
"meridian": {
"url": "https://www.meridian4pm.com/api/mcp"
}
}
}Check Cursor’s current documentation if the settings screen looks different.
Any tool that supports remote MCP servers over HTTP can connect. If it cannot sign in with OAuth, use a personal access token.
Authorization: Bearer <your token>Treat a token like a password. You can revoke it in Settings at any time.
What your AI tool can do
List your projects, open one for its details, and see its milestones and due dates.
list_projectsget_projectlist_milestones
Read the risk register with scores and response plans, and see change requests and their status.
get_risk_registerlist_change_requests
Read a project’s saved status reports, including health, sections, forecast and blockers.
get_status_report
See your own time entries, weekly totals, and the projects and tags you can log against.
Admins can also read time across the whole organization.
list_time_entriesget_timesheet_summarylist_time_projectslist_time_tags
See who has and has not logged time for a week.
Admins only.
get_team_status
Look up people, with their job titles, disciplines and email addresses. Cost rates are never included.
list_resources
Ask what your own role in Meridian lets you see and do.
get_my_access
Gather the facts for a status report: milestones, open risks, changes, sprint goals and hours, plus the minimum health the data supports.
Admins and Scrum Masters.
get_status_report_context
Read the stakeholder register: who each person is, their role, type and linked projects.
list_stakeholdersget_stakeholder
Read sprint retrospectives and the action items that came out of them.
list_retrospectivesget_retrospectivelist_retro_action_items
Read who has access to which software, when reviews are due, and how an entry changed over time.
get_access_matrixget_access_review_statusget_access_entry_history
These need the separate writes switch, and a connection that was granted the matching permission.
Create a time entry or correct one of your own. Entries can be previewed first, and a retry never creates a duplicate.
Admins, Scrum Masters and Members. Needs the writes switch.
create_time_entryupdate_time_entry
Save a status report for a project. It is published straight away under your name, so the tool previews it with you first.
Admins and Scrum Masters. Needs the writes switch.
create_status_report
Example prompts
“Which of my projects have a milestone that is overdue?”
list_projectslist_milestones
“Summarise the open high-score risks on Atlas and tell me which have no response plan.”
get_risk_register
“Which change requests are still waiting for a decision?”
list_change_requests
“How many hours did I log last week, by project?”
get_timesheet_summary
“Log 90 minutes on Atlas yesterday for design review.”
create_time_entry
“Draft this week’s status report for Atlas and show it to me before you save it.”
get_status_report_contextcreate_status_report
“List the open retrospective action items for Atlas and who owns each one.”
list_retro_action_items
Security and control
Every request runs with your role and your project access. If you cannot see something in Meridian, your AI tool cannot either.
MCP is off until an admin turns it on, and writes are a second, separate switch. Switching either off takes effect on the next request.
Disconnect a connected app or revoke a token under Settings, Integrations, and it stops working straight away. Changing your password also disconnects your connections.
Meridian records each tool call, including refused ones and retries, against the token that made it.
A connection lasts up to 90 days, then your tool asks you to sign in again. Personal access tokens last the 30, 90 or 365 days you choose.
Writes can be previewed first, a retry never creates a duplicate, and edits refuse to overwrite a change made since you last looked.
Anonymous retrospective cards never show their author to your AI tool. Not even admins can see who wrote them; only the author sees their own cards.
Common questions
It is an address your AI tool connects to, using the Model Context Protocol, so it can read your Meridian project data and, if your admin allows it, log time and save status reports. Meridian makes no AI calls of its own for this: your AI tool asks, and Meridian answers.
Professional, Team and Enterprise. Workspaces on the free Starter plan cannot connect, and an admin on a paid plan has to switch MCP on before anyone can.
Not by default. Connections can read only until an admin allows writes. Even then, the only writes are logging and editing your own time entries and saving a status report. There are no delete tools and no way to approve or reject a change request.
Exactly what you can see in Meridian. The connection uses your role and project access. Viewers can read but not write, Guests cannot connect, and the team directory never includes cost rates.
Open Settings, then Integrations, and choose Disconnect on the connected app, or revoke a personal access token. It stops working immediately. An admin can also switch MCP off for the whole workspace.
Any tool that supports remote MCP servers over HTTP. This page has setup steps for Claude.ai and Claude Desktop, Claude Code, ChatGPT and Cursor, plus a personal access token route for everything else.
The MCP server is part of the paid plans. Compare them, or talk to us about your setup.