Skip to content
Scroll
Multi-channel AI agent platform

Voice. Email. Chat.
AI you don't need a developer to build.

Build a V-Rep. Fund a wallet. Go live in minutes.
One console, three channels. No contracts. No code.

Scroll

Three channels. One V-Rep.

Get the highlights
Voice V-Rep

Jeeves picks up
the phone.

9 tabs. Cartesia, ElevenLabs, OpenAI, Rime, or Deepgram voice. Per-minute wallet. SIP or Vonage. Live in minutes.

v-rep.jeeves · cartesia / benedict01:14
Chat V-Rep

Iris answers
on the site.

6 tabs. Embeddable widget. Same knowledge base. Hands off to voice mid-session. No script gymnastics.

When does the dividend pay out?
Q2 dividend record date is May 21, payment June 9. The full schedule is on the IR page.
Can you have someone call me?
Yes. Connecting you to Jeeves now.
v-rep.iris · embeddedsession 02:11
Swipe
Scroll for all three
Tab 1 / 9
Keep scrolling to build
The signature moment

Watch Jeeves get built.

portal.theconnexus.ai · v-reps · voice · v-rep
Identity
Name your V-Rep, write its welcome message, set its system prompt. No code. Just words.
Name
Jeeves
Timezone
America/New_York
Welcome message
"Good afternoon, you've reached investor relations. This is Jeeves. How can I help?"
System prompt
You are Jeeves, the voice agent for the IR desk… [ 412 / 4096 tokens ]
Reasoning
OpenRouter routes the call to the model. Defaults are conservative; tune temperature, tokens, top-p without touching code.
Provider
OpenRouter
Model
openai/gpt-5.1anthropic/sonnet-4.6google/gemini-2.5
Temperature
0.35
Max tokens
512
Listening
Deepgram nova-3, every English accent. Tune interruption sensitivity and active listening per V-Rep.
Engine
Deepgram nova-3
Language
English (all accents)
Interruption sensitivity
Active listening
PII redaction
Voice
Pick one of five providers per V-Rep. Cartesia, ElevenLabs, OpenAI, Rime, or Deepgram. Voice Lab for custom clones.
CartesiaBenedict
ElevenLabsAlaster
OpenAIOnyx
RimeReginald
DeepgramAura
Response speed
Background sound
Call behavior
Silence handling, hangup logic, voicemail detection. The pieces every voice deployment gets wrong.
Silence detection
Hangup threshold
8 seconds
Max call duration
12 minutes
Voicemail detection
Tools
Custom Functions, MCP Server registration. Jeeves can quote stocks, post to CRM, end the call, transfer to a human.
Custom functions
get_stock_quotepost_to_crmend_calltransfer_to_human
MCP Server
https://mcp.client.example/ir
Tool success rate
99.2% last 7d
Analytics
Every call ends with structured JSON. Summary, outcome, sentiment, action items. Delivered via webhook to your stack.
Post-call extraction
Webhook URL
https://hooks.client.example/athena/calls
Delivery report
2,847 success · 3 failed (HTTP 503, retried)
# sample payload { "call_id": "call_42-198", "sentiment": "positive", "outcome": "qualified", "duration_sec": 74, "pii_redacted": true }
Memory
Conversation memory keyed to caller phone number. The second call already knows who's calling.
Key by
caller_phonesession_id
Retention
90 days, then summarized + purged
Omnichannel
share with Penelope (email) + Iris (chat)
Deploy
Embed the call widget on any page with two lines of HTML. Same widget powers in-page calling + chat.
<script src="https://cdn.theconnexus.ai/v1/athena.js"></script> <athena-talk v-rep="jeeves" position="bottom-right"></athena-talk>
Status
● live on client.example/ir
Latency
142ms p95 · LLM, synth, transcriber stack
5-Minute Super-Migration

From Retell or Vapi. In under five.

connexus-cli · zsh · migrate-from-retell
Total elapsed time, 04:58. Zero downtime.
IR desk · live deployment

One IR team. Two V-Reps. Three channels.

Jeeves
Voice V-Rep · Cartesia / Benedict
● ON CALL
01:14
89
calls this quarter
0:42
avg pickup latency

"Same knowledge base. Same brand voice. Email and voice answered the same investor question, same way, in the same hour."

Investor relations, publicly traded technology company · deployed 2026-Q1 · tenant: ir-desk
Sovereign by default

Your data never touches a public model. Ever.

Multi-tenant by architecture. PII redacted before reasoning. Logged to an immutable audit trail. SOC 2 Type II audited 2026-Q1.

SOC 2 Type II HIPAA GDPR PII Redaction PHI Masking No Model Training
Scroll to see the math
The math

One wallet. One blended rate. One ledger.

Wallet · IR Desk
$1,247.00
$0.099 / min blended · last 7 days
Auto-recharge on
Top up to $500 when balance < $50
Recent activity 2026-05-14 · 14:32 EST
Vapi · platform + 4 invoices
$0.13 – 0.33 / min
LLM, STT, TTS, telephony billed separately
Retell · with add-ons
$0.07 – 0.31 / min
denoising, PII removal, AI QA stack on top
ATHENA · blended wallet
From $0.079 / min
4 tiers ($0.079–$0.149). One wallet, no line items.
ATHENA

Hear it now. Right now.

No signup. No waitlists.
ATHENA is on the other end of the line.

Talk to ATHENA Start for $10, get $20 in credit*

ATHENA is an AI voice agent, not a human. Starting a call requests microphone access; calls may be recorded and analyzed to operate and improve the service. See our Privacy Policy.

*Optional launch promotion. Fund your wallet with the $10 minimum and opt into the $20 promotional credit to start with a $30 spendable balance. Promotional credit applies to platform usage. Subject to the Terms of Use.