Connect your own AI (MCP keys)

You can connect your own Claude, ChatGPT, or any MCP client to HuntSales with an API key and have it run your CRM for you. This is separate from the built-in Copilot: here you bring your own AI and point it at HuntSales over MCP (Model Context Protocol). Full setup steps are at /docs/mcp.

How to connect: open Admin, then Developer, and create an API key. Every member can do this, the page is not admin-only. Give the key a name, copy it (it is shown once), and add it to your AI client. You can revoke your own keys any time from the same page.

Keys are personal and inherit your role. A key can do exactly what YOUR role can do in the app, no more: there is no scope picking at creation, and every API call re-checks your CURRENT role permissions. If an admin changes your role, all your keys follow immediately; if a member is removed from the team, their keys stop working entirely. Admins (superadmins) see every key in the workspace under Developer with its owner, and can revoke any of them. A superadmin's own keys can do everything.

What the role ceiling means in practice: searching prospects and revealing their emails/phones over the API needs the Live Search permission (reveals spend People Search credits). Adding People Search prospects needs Add Prospects to the CRM, and each add spends 2 People Search credits. Creating and updating contacts, moving pipeline stages, and building lists needs Edit Contacts. Building campaigns, enrolling contacts, and launching needs Campaigns, Email Setup & Replies. Managing webhook endpoints over the API needs Automations. Reading your contacts, pipeline, contact lists, connected mailboxes, and credit balances is available to every member's key, but if your role lacks View All Contacts, your key reads only the contacts assigned or co-assigned to you plus unassigned ones, exactly as in the app. Mailbox reads cover the address, provider, status, and daily limit only; credentials are never exposed.

Safety model: the tools that add, spend, or start real sending (adding prospects, creating a list from a filter, enrolling a list, launching a campaign, and resuming a paused or editing a live one) run a dry-run preview first and then wait for your explicit confirm before doing anything. Creating a draft campaign, adding contacts to a list, and editing a draft apply straight away because a draft never sends until you launch it. Revealing prospect contact details spends credits on each call over the API. Per-call caps apply: adding is capped at 500 prospects per call and enrolling individual contacts at 2,000 per call. Launching checks the campaign has at least one step, a sending mailbox, and enrolled contacts, then only flips it live; your normal mailbox warm-up, daily and monthly send caps, and send window still pace the actual sending exactly as they do in the app. Connecting a mailbox stays a manual step you do in Email Settings; the AI can list your connected mailboxes but never sees their credentials. Writes fire your automations and webhooks exactly as the same action in the app would, so moving a contact's stage over the API can trigger any automation you have set up for that stage change. Every write the AI makes is recorded in the audit log, and revoking the key cuts access immediately.

A typical end-to-end ask, if your role includes the adding and campaign permissions, is something like: find HR managers in Singapore in People Search, add 200 into a new list, draft a 2-step campaign on your connected mailbox, enrol the list, preview the launch, and launch it once you confirm.