MCP server

Connect your AI agent

Let Claude Code, Cursor or Claude Desktop read your NextConstraint workspace directly: the leak costing you the most, its fix, your alerts and what past fixes did. Ask it to fix the constraint, and it can pull the evidence and the fix prompt itself, then mark the fix shipped when you say it's live.

What your agent can do

get_statusFunnel health right now: the locked constraint and its re-check, each step priced when nothing is locked, and what every watchdog check concluded.
get_current_constraintThe leak, what it cost per month when it was picked, and the one fix to make.
list_alertsWatchdog alerts, open and recent, most expensive first.
get_alertOne alert in full: what moved, what it costs, and the likely causes around that day.
get_fix_promptThe fix prompt for the constraint or an alert, in the format for Claude Code, Cursor, Lovable / Bolt / v0 or ChatGPT.
get_historyPast constraints and how each fix turned out.
mark_fix_doneRecords in NextConstraint that a fix shipped, which starts its re-check. The only tool that writes anything.

Read-only except mark_fix_done, which only changes NextConstraint's own record. Nothing here can touch your code, payments, analytics or site. Results carry counts, amounts, rates and dates, never your customers' names or emails. Text from outside NextConstraint, like commit messages and page text, is marked as data so your agent doesn't follow instructions hidden in it.

Create a token

In NextConstraint, open Settings → Connect your AI agent, name the token after where you'll use it (for example "Laptop · Claude Code") and create it. It starts with ncm_ and is shown once, so copy it then. The MCP server is part of the paid plan.

Add it to your agent

Claude Code

Run this in a terminal, with your token in place of ncm_…:

claude mcp add --transport http nextconstraint https://nextconstraint.app/mcp \
  --header "Authorization: Bearer ncm_…"

Cursor

Add this to ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "nextconstraint": {
      "url": "https://nextconstraint.app/mcp",
      "headers": { "Authorization": "Bearer ncm_…" }
    }
  }
}

Claude Desktop and claude.ai

Open Settings → Connectors → Add custom connector. Enter the URL https://nextconstraint.app/mcp, choose No sign-in, and add a request header named Authorization with the value Bearer ncm_… (with the space). If your dialog has no request headers section yet, use Claude Code instead.

Ask it

"What's my NextConstraint constraint, and what's the fix?"
"Get the fix prompt for Claude Code and make the fix on a new branch."
"Is anything broken in my funnel this week?"
"The fix is live, mark it done in NextConstraint."

Your agent asks before it calls mark_fix_done. Marking a fix done starts the same re-check as the button in the app.

Limits and revoking

If your agent asks you to sign in or authenticate, the token is missing, mistyped or revoked: there's no NextConstraint sign-in for agents, so create a new token and update the header.

Each token can make 60 tool calls a minute and 2,000 a day. It reads one workspace, the one it was created in. Revoke a token any time in Settings and it stops working at once; deleting your account revokes them all. NextConstraint keeps a hash of each token, never the token itself, plus which tool was called and when for 30 days. See the Privacy policy.