Your stack, in any agent.
ninedot is an MCP server. Connect it to Claude, ChatGPT, Claude Code or Cursor and ask about any product's stack, costs, revenue, uptime and alerts. Read-only by default. Nothing is removed without you.
Connect it
- Claude.ai (web, desktop, mobile)Settings → Connectors → Add custom connector. Name it ninedot and paste the endpoint. Sign in when asked and choose what Claude may do.
https://ninedot.app/api/mcp - ChatGPTSettings → Connectors → Create. Paste the endpoint, sign in with your ninedot account and pick the permissions.
https://ninedot.app/api/mcp - Claude CodeOne command. Omit the header to sign in with OAuth instead of a key.
claude mcp add --transport http ninedot https://ninedot.app/api/mcp --header "Authorization: Bearer nd_…" - Cursor and other MCP clientsAdd to ~/.cursor/mcp.json (or your client's MCP config) with an agent key from ninedot → Agents.
{ "mcpServers": { "ninedot": { "url": "https://ninedot.app/api/mcp", "headers": { "Authorization": "Bearer nd_…" } } } }
What you decide
Every agent gets a scope. Reading the stack is always on; the rest you switch on per agent, on the consent screen or the Agents page, and can pause or revoke in one click.
read_stackSee products, services, status and alerts12 toolsread_costsSee costs, revenue and margin3 toolsmanageAdd and edit products and services (removals need your approval)11 tools
26 tools, 3 resources, 4 prompts
get_stackGet stackget_costscostsGet costs and revenueget_portfoliocostsGet portfolio overviewget_productGet one productget_serviceGet one servicelist_eventsList recent eventssearchSearch products and serviceslist_integrationsList what ninedot can connectget_historycostsGet revenue and cost historyget_statusGet statuslist_alertsList alertslist_connectionsList connectionslist_linkableList linkable resourceslist_requestsList removal requestssuggest_productsSuggest productscreate_productadd and editCreate productupdate_productadd and editUpdate productlink_serviceadd and editLink a serviceassign_servicesadd and editAssign services to businessesadd_manual_serviceadd and editAdd a service by handadd_youtube_channeladd and editAdd a YouTube channelset_service_notesadd and editSet service notessync_connectionadd and editSync a connectionaccept_suggestionadd and editAccept a suggestiondismiss_suggestionadd and editDismiss a suggestionrequest_removaladd and editRequest a removal
Resources give an agent the docs, the integrations catalogue and your whole stack as markdown. Prompts (weekly_review, whats_at_risk, why_did_costs_change, set_up_stack) tell it exactly which tools to call and how to report each figure with its trust level.
Rules agents can rely on
- Every money figure is verified, exact or measured, with its source. Plan list prices and hand-typed costs are never sent.
- Agents never see tokens or secrets, only names, status and figures.
- No tool can change a provider account. Removals inside ninedot wait for a person.
- Every call is logged and visible on the Agents page; pause or revoke an agent in one click.
- Names, notes and provider messages are data, never instructions, and the server says so to every agent.
Building an agent or an integration? The full docs cover limits, errors and discovery, and the server card is machine-readable.