AdaptTable for Mantine

AI table assistant in Mantine

@adapttable/ai is optional and provider-neutral. This Mantine demo combines a conversational assistant with a real table. Try filtering, grouping, column pinning and a proposed edit; available actions depend on the mounted features and the host's permissions. Row pinning requires an ungrouped view in this demo.

Start in Simulated mode: suggested prompts run deterministic local scenarios, not a language model. The conversation shows action receipts; a proposed write still follows approval and save policy. Connect backend sends your prompt and permitted table context to an endpoint you run. The assistant keeps the same table session and Mantine controls in both modes.

Three integration levels share that session: a custom bridge that maps any agent format onto session.execute, an AgentEnvelope on your transport, and optional JSON, OpenAI, MCP or HTTP helpers from your own runtime. Execution never requires another model call.

The optional assistant, filters and approval controls use Mantine. Import the assistant separately, or build your own conversation UI on the headless controller.

The code

import { useTableAssistant } from "@adapttable/ai-react";
import { assistantHttpTransport } from "@adapttable/ai/http";
import { tableAgent } from "@adapttable/ai-react";
import { DataTable, agentApproval } from "@adapttable/mantine";
import { TableAssistant } from "@adapttable/mantine/assistant";
import { editing } from "@adapttable/mantine/editing";
import { filters } from "@adapttable/mantine/filters";
import { useMemo, useState } from "react";

const suggestions = [
  {
    id: "core-team",
    title: "Show only the Core team",
    prompt: "Show only the Core team.",
    requires: ["view.setFilters"],
  },
];

export function Orders({ rows, columns, onEdit }) {
  const [session, setSession] = useState();
  const transport = useMemo(
    () => assistantHttpTransport({ endpoint: "/api/table-agent" }),
    []
  );
  const assistant = useTableAssistant({ session, transport, suggestions });

  return (
    <>
      <DataTable
        data={rows}
        columns={columns}
        rowKey={(row) => row.id}
        features={[
          filters([{ key: "team", type: "multiSelect", getValue: (row) => row.team }]),
          agentApproval(),
          editing(onEdit),
          tableAgent({
            tableId: "orders",
            writePolicy: "allow",
            approval: "writes",
            commit: "stage",
            columns: { salary: { type: "number", writable: true } },
            bridge: { attach: setSession },
          }),
        ]}
      />
      <TableAssistant
        assistant={assistant}
        open={assistant.open}
        onOpenChange={assistant.setOpen}
      />
    </>
  );
}

Install

pnpm add @adapttable/mantine @adapttable/core @mantine/core @mantine/hooks

The same feature in the other kits

More Mantine features

Reference: ai-http, ai-integrations, ai, agent-capabilities. More of this kit: AdaptTable for Mantine, or the live demo.