AI Under Governance

The platform's rule for AI is simple: an agent gets exactly the access its user has — never more — and every action is audited. AI here is not a bolt-on chat window; it operates through the same typed, permission-checked surface as the UI and the API.

Ask, and get answers with charts

Ask in natural language — "top ten opportunities by amount in the East region this quarter?" — and get the result as data and as a chart, inline. No SQL, no export-to-spreadsheet detour.

Let AI do the work

Conversationally create and update records, run actions, and drive processes: "convert this lead; the account name is Acme Ltd." Risky operations ask for confirmation first, or can be routed into a human-review inbox so AI proposes and people approve.

Knowledge answers

Connect product sheets, playbooks, and policy docs, and the assistant answers from your knowledge — with the same permission filtering on what each person may retrieve.

Bring your own agent (MCP)

Every deployment exposes its objects and permitted actions as an MCP server — on by default at /api/v1/mcp. Point Claude or any MCP client at your app and that agent can inspect and operate it, under the signed-in user's permissions, with every call audited. Your app is AI-operable the day it ships, for free.

In HotCRM: two copilots ship ready — a sales copilot (lead scoring, customer-360 summaries, email drafting, revenue forecasting) and a service copilot (case triage, live-data Q&A) — built from six reusable AI skills. Swap the skills to retarget them at your own domain.

The conversational UI is part of the commercial runtime; the MCP surface — bring-your-own-agent — ships in open source.

For developers: agents, skills, and MCP configuration in AI.