The AI-native CRM you can run on your own server
AI-native means the assistant and the agents are part of the CRM, not a plugin on top. Here is what to look for, and how Relaticle does each of it.
What AI-native actually means
Three things have to be true of the CRM itself, before any model is involved.
The AI reads the same records you do
Companies, people, opportunities, notes, and every custom field are one data model that your team and your agents share. No export, no sidecar copy.
Agents get a door, and the door has a lock
A standard tool interface lets any agent work in the CRM, and every write passes through the same authorization a person would.
You choose the model
Cloud or local, today or next year. A CRM that only works with one vendor's AI has made the choice for you.
Five checks before you call a CRM AI-native
Run these against any CRM, including this one. The right-hand column is how Relaticle answers each.
Can an outside agent work in it?
Claude, ChatGPT, or an agent you wrote should get first-class access to records, not a screen-scraper.
In Relaticle
A first-party MCP server and a REST API. Agents connected over MCP write directly, under the same permissions as the user whose token they use.
Does the AI see your fields?
Your custom fields, option lists, and relations are where the real data lives. An assistant that only sees the defaults is guessing.
In Relaticle
Every active custom field is readable and settable by the assistant and by agents, with option labels shown as text, not raw IDs.
Does the in-app assistant ask first?
An assistant that can write to your pipeline needs a review step, or one bad prompt rewrites a quarter of deals.
In Relaticle
Rela proposes every change as a card and waits. You approve or reject each record before anything is written.
Can it run where your data lives?
If the only way to get the AI is the vendor cloud, the CRM is not yours in any way that matters.
In Relaticle
Self-host the whole thing with Docker Compose. Bring your own key for Claude or GPT, or point it at a local model with Ollama.
Do you pay for AI per seat?
Per-seat AI add-ons punish the teams that adopt it most.
In Relaticle
One flat price for unlimited users. AI is metered by what you use, in credits, not by how many people log in.
Connecting an assistant you already use? Relaticle for Claude Relaticle for ChatGPT
AI-powered, agentic, AI-native: which is which
Vendors use the three interchangeably. They are not the same thing.
AI-powered CRM
A classic CRM with AI features attached: summaries, lead scoring, email drafts. Useful, but the AI sits beside the data model rather than inside it.
Agentic CRM
A CRM where agents take multi-step actions toward a goal: find the stale deals, draft the follow-ups, schedule the tasks. The question is who checks the work before it lands.
AI-native CRM
A CRM built so that both its own assistant and outside agents work on the data model itself, with permissions and review designed in from the start. Relaticle is built this way. Outside agents write through the MCP server under your permissions, and the built-in assistant is agentic within one rule: it asks before it writes.
The same AI, on your own server
The self-hosted install is the same codebase as the cloud, assistant and MCP server included. Bring your own API key for Claude or GPT, or run a local model with Ollama, and keep every request on infrastructure you control.
Questions about AI-native CRMs, answered
Run the five checks on Relaticle yourself
Free to start, no credit card required. Self-host it yourself whenever you want.