# Mnemix - Full Text Knowledge Bundle
Generated: 2026-06-24T00:00:00.000Z
Canonical: https://mnemix.ai

---

## Canonical Summary

Mnemix is the memory + real-world enrichment API for AI voice agents. Same-number caller memory and available enrichment help agents start informed, with sub-300ms voice recall as a design target. 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.

## What/Who/Price/How

- 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). 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/calls/end for post-call write-back and GET /v1/caller/{phone_number} for read-only caller profiles. Sub-300ms voice recall is a design target at the Cloudflare edge.

## Wedge

Mnemix is the memory API where callers can arrive with same-number memory and available enrichment joined to the agent's context. Real-world identity, intent, and history are shaped for voice workflows, with sub-300ms voice recall as a design target.

### Caller-ID enrichment, voice-native
Twilio Lookup + Trestle person/company resolution + Baylio call-intent fan out before the first audio packet. The agent's first response is already informed.

### Sub-300ms voice recall as an edge design target
Cloudflare Workers across 5+ regions run memory and enrichment in parallel, engineered to stay inside the conversational voice latency budget. Designed for sub-300ms voice recall.

### Bi-temporal session memory
Four timestamps per fact (valid_from, valid_to, observed_at, ingested_at). Built for voice flows where context shifts mid-call.

## Pricing

| Tier | Price | SLA |
|---|---|---|
| Hobby | $0 Hobby tier | Community Discord |
| Starter | Starter - contact hello@mnemix.ai for pricing (private beta) | Email · 48h |
| Pro | Pro - contact hello@mnemix.ai for pricing (private beta) | Email · 24h + Slack channel |
| Elite | Elite - contact hello@mnemix.ai for pricing (private beta) | Slack DM · 4h + dedicated CSM |

## Public API surface

Voice entry (primary for voice; not deprecated):
- POST /v1/recall_and_enrich: Primary cold-caller voice path; recalls memory, creates unknown contacts, and starts Trestle/Twilio/Baylio enrichment.
- POST /v1/calls/end: Persist completed call transcripts, outcomes, and terminal memory observations.
- GET /v1/caller/{phone_number}: Read an existing caller profile without auto-creating a contact.

## Integrations

### Bland
Bland.ai pathway integration with pre-call enrichment and structured task memory.
Quickstart: https://mnemix.ai/integrations/bland

## How Mnemix compares

### Mnemix vs Mem0

**Verdict:** Mem0's developer experience for chat is genuinely good: a memory-as-a-service layer you wrap around LangChain, CrewAI, AutoGen, or your own agent loop, with unmatched vector-store choice. The moment you swap stdin for a Vapi webhook, you're outside their lane — no phone-native entry path, no Twilio/Trestle/Baylio enrichment, and claimed/unverified LongMemEval figures include 49% independent/community replication and a 93.4% self-reported score. Choose Mnemix if you're building voice.

Background: The mindshare leader in bolt-on agent memory. claimed/unverified 54k-star snapshot, $24M raised, 21+ framework integrations, 19 vector backends — chat/session memory, not voice or caller enrichment.

| Dimension | Mem0 | Mnemix |
|---|---|---|
| GitHub stars | claimed/unverified snapshot: 54,251 | n/a (closed alpha) |
| Agent framework integrations | 21+ (LangChain, CrewAI, AutoGen, Mastra, …) | Voice platforms + MCP (Vapi, Retell, Bland, LiveKit, Twilio) |
| Funding | $24M (YC, Peak XV, Basis Set) | Bootstrapped |
| Voice integrations | ❌ | ✅ Twilio, Vapi, Retell, Bland |
| Caller-ID enrichment | ❌ | ✅ Twilio Lookup + Trestle + Baylio |
| Bi-temporal memory | ❌ | ✅ (4 timestamps per fact) |
| Edge runtime | ❌ Python server | ✅ Cloudflare Workers |
| Vector backends | 19 | 1 (Supabase pgvector) |
| Multilingual NER | ❌ English-only spaCy | ✅ libphonenumber + Trestle |
| LongMemEval (self-reported) | claimed/unverified: 93.4% | Published once measured |
| LongMemEval (independent) | claimed/unverified: 49% (community replication) | Published once measured |
| Compliance | Not asserted in public copy | No public compliance claim |
| Starter price | $19 to $249 cliff (13.1x) | Hobby $0; Starter+ contact sales |

Full comparison: https://mnemix.ai/compare/mem0-vs-mnemix

### Mnemix vs Zep

**Verdict:** Zep's bi-temporal graph is the strongest research-cited memory architecture for chat agents and Graphiti is genuinely impressive engineering. But voice is a community demo, not a first-class integration, and claimed/unverified memory benchmark figures include a 63.8% Mem0-paper measurement and a 71.2% self-reported score. Choose Mnemix if you're building voice.

Background: The research-cited choice. Bi-temporal knowledge graph (Graphiti), BYOC. Voice is a community demo only.

| Dimension | Zep | Mnemix |
|---|---|---|
| GitHub stars | claimed/unverified snapshot: 25,463 (graphiti) + 4,495 (zep) | n/a |
| Bi-temporal | ✅ (4 timestamps in graphiti_core/edges.py) | ✅ (session-scoped) |
| Voice integrations | 🟡 community demo only | ✅ Twilio, Vapi, Retell, Bland |
| Caller-ID enrichment | ❌ | ✅ Twilio Lookup + Trestle + Baylio |
| Edge runtime | ❌ Python on AWS | ✅ Cloudflare Workers |
| Claimed P95 retrieval | claimed/unverified: <200ms | designed for sub-300ms voice recall |
| LongMemEval (self) | claimed/unverified: 71.2% | Published once measured |
| LongMemEval (independent) | claimed/unverified: 63.8% (Mem0 paper) | Published once measured |
| Starter price | $25 to $125 (5x) | Hobby $0; Starter+ contact sales |

Full comparison: https://mnemix.ai/compare/zep-vs-mnemix

## FAQ

### What is Mnemix?
Mnemix is a memory + real-world enrichment API for AI voice agents. It joins identity, intent, and call history to your agent's memory at Cloudflare edge latency, with sub-300ms voice recall as a design target.

### Who is Mnemix for?
Developers building AI voice agents on platforms like Vapi, Retell AI, Bland AI, LiveKit, or directly on Twilio. Also chat developers who want phone-native identity.

### How is Mnemix different from Mem0?
Mem0 is a general-purpose bolt-on memory layer with strong mindshare, 21+ agent-framework integrations (LangChain, CrewAI, AutoGen, and others), and 19 vector backends. It extracts and retrieves chat facts across sessions — but does not resolve phone callers or fan out Twilio Lookup + Trestle + Baylio enrichment before a voice turn. Mnemix is voice-first with caller-ID enrichment built in, runs on Cloudflare Workers at the edge, and ships first-class integrations with Vapi, Retell, and Bland. Choose Mnemix if you're building voice. Choose Mem0 if you need a general-purpose memory layer with deep vector-store and framework choice.

### How is Mnemix different from Zep?
Zep + Graphiti is the research-cited bi-temporal knowledge graph for chat agents. Mnemix is voice-native with built-in caller-ID enrichment. Zep's voice support is a community demo; Mnemix ships first-class integrations with Vapi, Retell, Bland, and Twilio. Choose Mnemix if you're building voice.

### How is Mnemix different from Supermemory?
Supermemory and Mnemix share the same Cloudflare Workers stack. The difference is voice + enrichment: Mnemix bundles Twilio Lookup, Trestle, and Baylio call-intent into the memory primitives. Supermemory is general-purpose memory. Choose Mnemix if you're building voice.

### How is Mnemix different from Letta?
Claimed/unverified Letta voice-status snapshot: March 2026 HTTP 410 in letta/server/rest_api/routers/v1/voice.py. Mnemix is voice-first. Choose Mnemix if you're building voice.

### How is Mnemix different from LangMem?
LangMem is bound to LangGraph and is aimed at graph-backed agent workflows, not cold-caller voice recall. Mnemix is designed for sub-300ms voice recall with caller enrichment in the hot path. Choose Mnemix if you're building voice.

### How much does Mnemix cost?
Hobby is $0/month (free tier). Starter, Pro, and Elite tiers are in private beta — contact hello@mnemix.ai for pricing while we wire billing.

### Is there a free tier?
Yes. Hobby is $0/month, free forever. Attribution required (Powered by Mnemix). Sign up at https://mnemix.ai.

### What latency does Mnemix offer?
Mnemix is designed for sub-300ms voice recall at the Cloudflare edge. We publish measured numbers only once they are real — we would rather under-promise and ship honest results.

### Where does Mnemix run?
Cloudflare Workers (edge), with Durable Objects for session state, R2 for blob storage, KV for hot cache, and Supabase Postgres + pgvector for the durable memory store.

### Does Mnemix support the OpenAI Assistants API?
Yes — Mnemix is a migration target for voice teams. The OpenAI Assistants API is shutting down 2026-08-26, and Mnemix carries caller identity, enrichment, and memory across the move.

### Does Mnemix integrate with Vapi?
Yes. Mnemix is a plain HTTP API, so it works with Vapi today — call POST /v1/recall_and_enrich from your Vapi function-call handler and callers arrive already enriched. A dedicated step-by-step Vapi guide is on the way; the Bland guide at https://mnemix.ai/integrations/bland shows the same pattern.

### Does Mnemix integrate with Retell AI?
Yes. Mnemix works with Retell over plain HTTP today — call POST /v1/recall_and_enrich from your Retell webhook for caller-ID enrichment and conversation memory. A dedicated Retell guide is on the way; the Bland guide at https://mnemix.ai/integrations/bland shows the pattern.

### Does Mnemix integrate with Bland AI?
Yes. See https://mnemix.ai/integrations/bland.

### What benchmark scores does Mnemix have?
Mnemix does not publish benchmark numbers yet. Voice-grade memory benchmarks will ship only once they are independently measured — we would rather show real results than projected ones.

### Is Mnemix open source?
The enrichment SDK is MIT-licensed on GitHub. The memory layer is closed-source SaaS with self-host coming for Elite tier.

### Is Mnemix HIPAA-compliant?
Not as a present-tense guarantee. Mnemix is building toward SOC 2 and GDPR alignment, and HIPAA support with a BAA is on the roadmap — none are certified today. Email hello@mnemix.ai to talk through compliance timelines for your use case.

### Where is Mnemix data stored?
Cloudflare's global edge holds hot session state; durable storage runs on Supabase (US, Pro). Email hello@mnemix.ai about EU data-residency requirements.

### How do I get started with Mnemix?
Sign up at https://mnemix.ai for a free Hobby key and install the SDK (npm install @mnemix-ai/client). For cold voice callers, call POST /v1/recall_and_enrich before the first turn; use POST /v1/calls/end for post-call write-back and GET /v1/caller/{phone_number} for read-only caller profiles.

## Discovery surfaces

- https://mnemix.ai/llms.txt
- https://mnemix.ai/llms-full.txt
- https://mnemix.ai/agents.json
- https://mnemix.ai/.well-known/openapi.json
- https://mnemix.ai/api/agents/knowledge

Last updated: 2026-06-24