Installation
Installation
Mnemix meets your agent where it lives. There are three surfaces, and most teams use more than one:
- CLI — authenticate, look up callers, run recall demos, and inspect your API key status from your terminal.
- MCP server — connect Mnemix to any agent that speaks the Model Context Protocol (Claude Code does, natively) so caller memory surfaces as native tools.
- SDK / client — call the memory and enrichment APIs from your own application code.
Prerequisites
- Node.js 20.12+ for the CLI (the SDK alone works on Node 18+).
- A Mnemix API key. Hobby is $0; paid tiers are Contact sales at mnemix.ai. Your tenant is derived from the key — there's nothing else to configure.
Set it once in your shell:
export MNEMIX_API_KEY="sk_live_…"
1. CLI
npm install -g @mnemix-ai/cli
mnemix login # paste your key when prompted
Verify your credentials resolve:
mnemix whoami
# → key: sk_live_…xxxx
# base: https://mcp.mnemix.ai
# source: env
# key valid ✓
Available commands: demo (offline sample), login, recall <phone>, caller <phone>, whoami. Try a live recall:
mnemix recall +15551234567
2. MCP server (recommended for agents)
The MCP server is published as @mnemix-ai/mcp-server. It exposes Mnemix to any MCP-capable agent as native memory tools — lookup, save, search, history, enrich, and update.
Configure Claude Code (or any MCP client)
{
"mcpServers": {
"mnemix": {
"command": "npx",
"args": ["-y", "@mnemix-ai/mcp-server"],
"env": {
"MNEMIX_API_KEY": "${MNEMIX_API_KEY}"
}
}
}
}
MNEMIX_API_KEY is required; MNEMIX_API_URL is optional (defaults to https://mcp.mnemix.ai). Restart your client and the mnemix tools appear.
Prefer a hosted connection? Point your MCP client at the remote HTTP transport instead. Note the "type": "http" transport discriminator — it is required for an HTTP MCP server entry:
{
"mcpServers": {
"mnemix": {
"type": "http",
"url": "https://mcp.mnemix.ai/mcp",
"headers": {
"Authorization": "Bearer ${MNEMIX_API_KEY}"
}
}
}
}
3. SDK / client
npm install @mnemix-ai/client
import { Mnemix } from "@mnemix-ai/client";
const mnemix = new Mnemix({ apiKey: process.env.MNEMIX_API_KEY! });
// The voice hot path (live) — recall + enrich a caller at ring time
const recall = await mnemix.recallAndEnrich({
phone_number: "+15551234567",
trigger: "answered",
session_id: "call_abc123",
});
// Post-call write-back — persist the completed call outcome
await mnemix.callsEnd({
phone_number: "+15551234567",
session_id: "call_abc123",
transcript: [
{ role: "user", text: "I'd like a callback tomorrow morning.", ts_ms: 1719861600000 },
{ role: "agent", text: "Done — we'll call tomorrow morning.", ts_ms: 1719861608000 },
],
duration_s: 84,
outcome: "callback_requested",
});
Integration kits for common voice stacks (@mnemix-ai/bland-kit, @mnemix-ai/vapi-kit) live in the same repo; check the registry for current publish status.
Endpoints at a glance
| Surface | Base URL |
| :--- | :--- |
| Everything — REST v1 + MCP transport | https://mcp.mnemix.ai (MCP at /mcp) |
One host. The frozen public v1 surface is POST /v1/recall_and_enrich, POST /v1/calls/end, and GET /v1/caller/{phone_number}.
Verify your install
mnemix whoami # key valid ✓
mnemix recall +15551234567
If the recall card renders, you're ready for the Quickstart.