Intempt CLI + MCP
The Intempt platform you talk to.
Ask it in plain language from Claude, ChatGPT, or your coding agent - or script it directly from the terminal. 202 real operations across 12 domains, no export step.
What you can do
Which accounts have power users who went quiet this month?
Resolves against the real accounts and users domains - no export, no spreadsheet.
Create a segment of trial users who haven't converted in 14 days.
One call against the segments domain builds it live, ready to route into a journey.
Get the funnel drop-off for our signup flow this quarter.
Pulls real funnel step analytics - the same data your dashboards already show, without opening one.
Summarize my last three meetings with Acme Corp.
Reads real meeting summaries and activity - the meetings domain, not a guess.
Draft outreach for the accounts that just hit power-user status.
Uses the same outreach drafting Blu runs in-app, callable directly from your agent.
List every journey with a live A/B experiment right now.
journeys is the largest domain in the registry (43 real operations) - experiments, schedules, and drafts included.
What's our brand voice guidance for outbound emails?
Reads real brand-kit and persona data so generated copy actually matches your brand, not a generic tone.
Pull the deal activity for every open deal over $50k.
Real deal and activity data from the deals domain - no CRM export required.
Get started
Pick how you want to use it, paste the matching prompt into your agent.
For asking questions from Claude, ChatGPT, Claude Code, Codex, or any MCP client - the identical server either way, only the framing of what you ask changes.
setup prompt - paste into Claude Code
You are helping me connect Intempt as an MCP server. It exposes
the @intempt/commands registry (202 real operations across 12 domains -
accounts, users, segments, events, deals, journeys, meetings, experiences,
analytics, blu-chat, designer) as MCP tools I can call by asking in plain
language. The package isn't on the public npm registry yet, so this builds
it locally from source - same server, just not a bare npx install for now.
Work one step at a time, run commands for me where you can, and wait for my
confirmation before moving on.
1. Clone https://github.com/intempt/cli, then in packages/mcp run
"bun install" and "bun run build". This produces packages/mcp/dist/index.js.
2. Register the server using that local build's absolute path:
- In Claude Code: claude mcp add intempt -- node <absolute path to>/packages/mcp/dist/index.js
- For Claude Desktop or another MCP client, add this under mcpServers in
its config, and tell me where that config file lives on my machine:
{"command": "node", "args": ["<absolute path to>/packages/mcp/dist/index.js"],
"env": {"INTEMPT_ORG": "<my org slug>", "INTEMPT_PROJECT": "<my project slug>"}}
3. Confirm the intempt tools load, then say "log me in to Intempt" - a
browser window opens for device-code login, no terminal step needed.
4. Show me one real call working, for example listing accounts or pulling a
recent meeting summary.
If a step fails, show me the exact error and fix it with me before moving on.FAQ
What can it actually access?
Whatever your INTEMPT_ORG and INTEMPT_PROJECT scope it to - the same org/project boundary that applies everywhere else in Intempt. A compromised MCP config can't see other orgs or projects.
Do I need to be a developer to use this?
To talk to it through Claude or another MCP client, no - you install it once (or ask your agent to do it for you) and then just ask questions in plain language. The CLI mode is for people who want scriptable, terminal-native output.
How does authentication work?
Device-code login runs from inside the chat itself - say "log me in to Intempt" and approve in your browser, no terminal needed. CI and automation use a static token instead.
Is this the same tool whether I use ChatGPT, Claude Code, or Claude Desktop?
Yes. All of them connect to the identical MCP server over the same protocol - the only real difference is where you're asking from, not what you get access to.
What's the difference between MCP mode and the CLI?
MCP mode is for asking questions inside a chat or coding agent. The CLI is for scripting and terminal use - every command supports --json output for piping into other tools.
Where do my credentials live?
Locally, in your own environment or MCP client config - INTEMPT_ORG, INTEMPT_PROJECT, and your auth token never pass through a third-party server.
Is this an official Intempt product?
Yes - built and maintained by Intempt, not a community integration.