Bring your own AI
Connect the AI you already use — any of them
The old idea of a CRM was "our built-in assistant, billed in credits". AKM Advisor drops that. Your CRM becomes a system your own AI works in: Claude, ChatGPT, Gemini, Copilot, an open-source model on your laptop or a bot you wrote — anything that can read a link and call an HTTP API. You give the agent a link with a key and a protocol; it does the rest. The agent reads your business brief, works inside the CRM with the permissions you set, and writes back what it did. You pay your AI provider; we don't mark it up.
- 1One link, no tokens to paste. Copy the connection link from Agents Connection and hand it to the agent. It downloads the protocol itself and checks for updates before each session.
- 2Permissions per module. Leads, deals, projects, forms, finance, HR, products, knowledge — none, read or write for each connection. Give a browser agent lead-write only and keep it out of invoices.
- 3A business brief, not a prompt. On first connection the agent interviews you (25 questions, in your language) and stores the brief in the system with revision history. Every agent you connect later works from the same facts.
- 4Agents take turns. Several agents can work one task in relay — a chat agent drafts, a browser agent researches — with one baton, a shared thread and a watchdog that frees a stuck turn.
$ curl https://app.akm-advisor.com/api/v1/agent/crm/ping \
-H "X-Agent-Key: ••••••••"
{
"ok": true,
"agent_name": "sales-assistant",
"modules": {
"leads": "write",
"deals": "write",
"projects": "read",
"forms": "write",
"finance": "none",
"knowledge": "write"
}
}
$ curl -X POST .../agent/crm/cards/lead \
-H "X-Agent-Key: ••••••••" -d '{
"name": "Harbor Engineering",
"source": "browser:linkedin",
"contacts": [{ "email": "dana@harbor.example" }],
"original_data": "…full profile text…"
}'
