mnemix@0.2.0 / public-beta

The memory layer for voice AI agents.

Voice agents can't wait. Mnemix answers "who is this caller?" with full context, enrichment, and history — designed for sub-300ms voice recall.

Powering voice agents at Baylio.io

Live API demo

What you get back.

One lookup gives the voice agent a contact profile, enrichment sources, AI summary, and recent history in the same response shape your call logic can consume immediately.

curl https://api.mnemix.com/v1/lookup/+15551234567 \
  -H "Authorization: Bearer sk_live_..."

HTTP/1.1 200 OK
x-mnemix-cache: hit
x-mnemix-latency: 4ms

{
  "known": true,
  "contact": {
    "id": "ct_7Br4...",
    "name": "Mike Reynolds",
    "company": "Reynolds Auto Repair",
    "phones": ["+15551234567"],
    "relationship_status": "active",
    "summary": "Repeat customer with a fleet account. Last called about a brake inspection and prefers morning appointments.",
    "sentiment": "positive"
  },
  "enrichment": {
    "trestle": {
      "caller_name": "Mike Reynolds",
      "line_type": "mobile"
    },
    "twilio": {
      "carrier": "Verizon",
      "valid": true
    },
    "google_business": {
      "name": "Reynolds Auto Repair",
      "rating": 4.6,
      "reviews": 183
    }
  },
  "recent_interactions": [
    {
      "summary": "Called about fleet brake inspection",
      "occurred_at": "2026-04-12T14:30:00Z"
    },
    {
      "summary": "Asked for early drop-off hours",
      "occurred_at": "2026-03-29T11:05:00Z"
    }
  ]
}

name + company

Trestle

Resolved caller identity and business profile before the agent speaks.

Google rating

Google Business

Adds local business reputation and review count to the profile.

sentiment

Claude Haiku

Summarizes relationship tone from previous conversations.

recent interactions

Mnemix DB

Returns complete interaction history without transcript stuffing.

<300ms

designed for voice recall p99

edge

Cloudflare Workers + Hyperdrive

soon

public benchmarks

1 call

full context

Enrichment waterfall

Cold numbers become caller context.

Mnemix runs the slow work once, writes the memory, then returns subsequent lookups from cache — designed for sub-300ms voice recall.

Cold lookup

+1 555 123 4567
Name
Company
Rating
Summary
History

After Mnemix

+1 555 123 4567
NameMike ReynoldsTrestle
CompanyReynolds Auto RepairTrestle
CarrierVerizon mobileTwilio
Rating4.6 from 183 reviewsGoogle
SummaryFleet account, morning appointmentsClaude
HistoryBrake inspection, early drop-offMnemix

Phone in

0ms

Cache check

2ms

Trestle

180ms

Twilio

90ms

Google Business

210ms

Claude Haiku summary

400ms

Written + returned

cold path

Cold path enriches once; every repeat lookup is served from memory cache — designed for sub-300ms voice recall.