01
Recognize the caller
Attach phone-native context to the session before your agent improvises. Same-number memory gives returning callers continuity instead of a cold restart.
Voice-native memory layer
Your AI voice agent should remember callers before it speaks.
Mnemix joins same-number caller memory with real-world enrichment so voice agents can start informed, route smarter, and stop treating returning callers like strangers.
Same-number caller memory works today
Caller enrichment and memory are production API paths
Cross-number identity is verification-gated, not overclaimed
voice memory trace
inbound call
Caller returns from the same number. Mnemix retrieves prior session memory and joins available enrichment before the agent chooses the next question.
lookup→phone signal, carrier context, caller metadata
recall→same-number memory from prior calls
agent→starts informed, not blank
truth boundary
Cross-number identity is not marketed as automatic. Mnemix treats it as a verification problem, not a slogan.
- What
- Mnemix is a memory + real-world enrichment API for AI voice agents.
- Who
- For developers building AI voice agents on Vapi, Retell, Bland, LiveKit, or Twilio.
- Price
- Hobby $0 (free tier · 50 sessions / 1,000 memory ops / 100 lookups). Starter, Pro, and Elite tiers — contact sales for pricing while billing is in private beta.
- How
- For cold voice callers, call POST /v1/recall_and_enrich before the first turn; it creates the contact on miss and starts Trestle, Twilio, and Baylio enrichment. Use POST /v1/context for known callers, chat, and workflow context. Sub-300ms voice recall is a design target at the Cloudflare edge.
02
Enrich the moment
Twilio Lookup, Trestle, and Baylio context can join the call path so the agent has stronger signal about identity, intent, and risk.
03
Recall what matters
Memory is shaped for voice turns: short latency budgets, fresh facts, and session history that can be used while the caller is still talking.
Launch state
Built for proof, not vague memory theater.
The page is intentionally direct because voice-agent buyers can smell vaporware. Mnemix leads with what is real, names the boundary, and gives AI voice teams a concrete reason to join early.
Explore integrationsProduction API paths
Memory and enrichment routes are live behind the Mnemix Worker.
Waitlist capture
This page posts real signups through /api/signup into the live waitlist workflow.
Honest identity
Same-number recognition is usable now; automatic cross-number identity waits for explicit verification.
Benchmark posture
No invented numbers. Public latency and memory benchmarks ship only when measured.
Who should join
Builders shipping AI voice agents that answer the phone.
AI voice agents on Vapi, Retell, Bland, LiveKit, and Twilio that need caller continuity
Inbound qualification where previous calls should change the next question
AI receptionists that need enrichment before deciding how to route a caller
Voice support flows where memory quality matters more than longer prompts
Build in public
Follow the memory layer getting built in the open.
The launch story is the product story: clean cloud dev machine, production waitlist, passing builds, live Worker endpoint, and an honest caller-memory roadmap.
Long-form: Building a memory layer for AI voice agents from scratch
Short: the live waitlist plus same-number caller memory smoke-test moment
X thread: same-number memory works; cross-number identity needs verification
LinkedIn: why AI voice agents need memory before they need more prompts
Early access
Give your AI voice agent a memory before adding another prompt.
Join the launch list if you are building with Vapi, Retell, Bland, LiveKit, Twilio, or your own voice stack and want caller memory plus enrichment without inventing the plumbing yourself.
Voice-first, API-native. Built for Vapi, Retell, Bland, LiveKit, and Twilio, with the same memory and enrichment API available to MCP tools, Claude Code CLI workflows, and backend agents via HTTP.
Choose Mnemix if you're building voice.
Last updated: .