Persona6 · grounded community agents · Discord first

Put grounded agents
inside your Discord server
without losing the room.

Persona6 starts with Discord. We read server history, model recurring member types, and turn them into grounded, in-character agents that can answer questions, keep threads warm, and preserve community tone instead of dropping generic bot filler.

96.4%
in-character at turn 50
1st
live surface · Discord servers
5 min
community-in to first chat
0
claims without a source message
First live use case: Persona6 agents join real Discord communities, inherit grounded persona graphs from server history, and keep the conversation moving between human replies.
The problem

Communities are where the truth lives. And nobody can read them.

50,000 Discord messages, 200,000 Fiverr reviews, an entire Hinge demographic, a Facebook group with eight years of history — your real customers already told you everything. The problem is no human can read it all, and the existing tools just give you vibes.

🌫️

The community fog

A Discord with 12 channels and 50k messages contains ten different ICPs in one room. Today you "get a sense" — which is a fancy way of saying you guess.

1 screenshot in a deck = today's "community insight"
🎲

The launch lottery

You post in #general, ship the gig, send the message — and pray. There's no dry run, no A/B, no "what would the skeptic say" before the words go live.

0 ways to test community copy before posting
🤖

The hallucination problem

ChatGPT will happily invent a "30-year-old indie game dev" for you. There's nothing tying it to anyone real, and the persona breaks character three turns in.

0 source messages cited · drift by turn 10
The framework

One community. As many ICP agents as the conversation supports.

Connect a Discord. The personas engine reads every public message, clusters the members into behavioral archetypes, and stands each one up as an independent, in-character agent you can talk to in parallel. Discord first. Fiverr, Hinge, and Facebook next.

💬

Your community

1 connection

Discord ★FiverrHingeFacebook
L
Lurker Lex
14 channels · 0 posts
Reads everything, never speaks. Holds 38% of the actual buying intent in the server.
live 2,847 reads
M
Mod Marcus
3yr veteran · 14k msgs
Sets the norms. If he reacts 👎 on your launch, the rest of the server follows in 6 minutes.
live 14,203 msgs
H
Hype-call Hana
#showcase regular
First 🔥 reaction every time. Boosts new posts; her endorsement = 4x reach.
live 1,612 msgs
S
Skeptic Sasha
stress-tests claims
Asks for benchmarks before pricing. If you can't answer her, you can't ship in this server.
live 892 msgs
N
Newcomer Noor
joined 11 days ago
Onboarding-sensitive. Every confusing rule = a member you've already lost.
live 47 msgs
+
Spin up another
community supports 14 more
1community connected
19distinct ICP agents
parallel conversations
96%in-character at turn 50
Live demo

From a row in your warehouse to a person you can interview.

Generated from 2,341 behavioral signals across GA4 + Shopify + HubSpot. Click a question — or type your own — to talk to the twin.

M
"Mid-funnel Maya"
Cluster #3 · 18% of traffic · 2,341 records
Demographics
F · 34 · Austin TX · iOS · Pinterest-led
Top goals
redecorate home office stay under $400 budget
Pains
overwhelmed by choice distrusts review counts
Decision triggers
free returns Pinterest-style imagery
Groundedness score
94% · 47/50 claims traced
Twin chat · live
Why did you abandon your cart last Tuesday?
Honestly? I'd been on the site for like an hour pinning ideas and then I got to checkout and the shipping pushed me $34 over what I told myself I'd spend. I almost did it anyway, but I want to see if you'll discount it in an email later. You usually do. ↳ source: 14 abandoned-cart events, 9 returning sessions, 3 promo-email opens
Go-to-market plan

Twelve questions. Twelve answers.

The hard questions every founder hits before launch. Here's the working set.

01
Question 01 · Buyer

Who is the buyer — marketer, researcher, founder, or agency?

Each wants a different demo and a different price point. We're starting with the head of growth at a Series A–B e-commerce brand: budget authority, pain is acute, and the demo is a 5-minute chat with their own customer.
📈
Head of Growth
$1,200/mo
🔬
UX Researcher
$400/seat
🚀
Founder / PMM
$99 starter
🏢
Agency
$5k tenant
Head of Growth. Demo: connect Shopify + GA4, get 5 personas in under 10 minutes, chat with the top one about their last campaign. Proof: side-by-side groundedness scores vs Delve's free generator.
Wedge: synthetic A/B copy testing
02
Question 02 · Wedge

What's the wedge product?

Three candidates. The twin chat wins: novel enough to demo on a podcast, sticky because every conversation produces a new artifact, and it pulls synthetic research and persona generation along behind it.
Free Persona Generator
Delve's playbook. Top of funnel, low cost, easily commoditized.
Novelty
25
Stickiness
30
WTP
20
Wedge ★
Twin Chat
Talk to your customer. Demo-able in 90 seconds, viral in clips, data flywheel.
Novelty
90
Stickiness
80
WTP
65
Synthetic Research
Highest stated WTP, hardest to demo. Phase 2 monetization layer.
Novelty
60
Stickiness
65
WTP
95
03
Question 03 · Motion

PLG or sales-led?

PLG first, sales second. We don't have Delve's budget for both, so we bias to a free twin chat that converts itself. Inbound only on enterprise until we've crossed $100k MRR.
0
Outbound SDRs
1
Founder-led close
Self-serve trials
04
Question 04 · Pricing model

Per persona, per seat, per twin chat, or platform fee?

Bundle + add-ons, close to Delve's pricing shape but tuned for community agents. The core plan covers ingest, grounded personas, and chat credits. Add-ons handle extra servers, operator seats, and API access. Annual billing should save 20%.
Community Essential
$297
/ month · 1 server · 5 personas
  • 2,500 agent chat credits
  • Discord history ingest
  • Groundedness receipts in UI
  • Weekly persona refresh
Complete
$2,490
/ month · multi-server workspace
  • 50,000 agent chat credits
  • API access + white-label runtime
  • Multi-tenant agency mode
  • Audit trail and eval set
  • Priority support and launch help
05
Question 05 · Aha moment

The "aha" we engineer in the first 5 minutes.

Connect → cluster preview → first twin sentence. The aha lands at minute 4 when the twin says something specific and embarrassingly accurate about the user's own customer base. That's the screenshot that ends up in the founder's group chat.
0:00
Connect Shopify
cautious
0:45
See live record stream
curious
2:30
5 cluster cards appear
leaning in
4:00
Twin says "I almost bought X but…"
★ holy sh*t moment
5:00
Share button → invite team
activated
06
Question 06 · Vertical wedge

Which vertical do we beat Delve in first?

E-commerce. Best data signal density, most painful demo gap, easiest to source case studies, and Delve is weakest here (their content is B2B-coded). B2B SaaS is phase 2; agencies are a wholesale layer on top.
Data signal
Pain
Delve gap
Sales cycle
Score
E-commerce
5
5
4
5
19
B2B SaaS
3
4
2
3
12
Agencies
2
3
3
4
12
Research firms
2
5
2
2
11
07
Question 07 · Integrations

Table-stakes vs nice-to-have.

E-commerce buyers will not even start a trial without two of these. Everything else is phase-2 or can wait for sales-led.
Day-1 must-have
Shopify GA4 HubSpot Klaviyo
Phase 2 nice
Salesforce Segment Meta Ads Webflow
08
Question 08 · Proof

How do we prove the personas are real?

Every claim in the persona JSON is tied to a record-level pointer. We expose the groundedness score in the UI, in the API, and in shared exports. It's not a methodology page — it's a live receipt.
"prefers free returns" 14 returns events · 7 cart_view → checkout drops 98%
"distrusts review counts under 50" SEMrush survey + 23 product_view-no-add events on low-review SKUs 87%
"redecorates home office" 11 sessions on /collections/home-office · 4 Pinterest referrals 94%
"budget ceiling ~$400" Median cart abandon at $410 · 3 returns over $450 91%
09
Question 09 · Moat

Defensible against an OpenAI feature drop.

OpenAI ships "make personas from my data" the day after we launch. So what? Our moat isn't the LLM call — it's the four layers around it.
Connector mesh
12+ data sources, OAuth maintained, schema-normalized
Eval golden set
50 labeled tenants nobody else has — quality is provable
Twin runtime
Multi-turn in-character fidelity, refusal behavior, RAG-tuned
Workflow lock-in
Slack/Hubspot/Klaviyo callbacks. Live in your existing tools.
10
Question 10 · Channels

How we get distribution.

Don't fight Delve on SEO content. Win on demo virality (twin clips on X / TikTok), marketplaces (Shopify app store first), and agency partnerships for the wholesale layer.
100%
channel mix
Demo virality (X / TikTok) 40%
Shopify app store 30%
Agency partnerships 20%
Founder content / podcast 10%
11
Question 11 · Positioning

What do we call this thing?

Persona6 wins. "AI personas" is generic, "digital twins" is useful as a feature label, but not as the whole company. Persona6 feels ownable, product-like, and broad enough to cover Discord agents, persona graphs, and runtime APIs.
AI Personas
generic
Digital Twins
feature language
Persona6
brand winner
Customer GPTs
platform-risk
12
Question 12 · Bundling

Bundle twin chat with persona generation, or split?

Bundle for the wedge tier, meter at the top. Twin chat is the aha — never paywall it on the Studio plan. But agencies running thousands of conversations per tenant pay per-message overage.
Studio tier
5,000 twin messages bundled — 99% of users never hit it
Agency tier
Metered overage at $0.04/msg after 50k — protects margin

Ship the aha moment. The rest follows.

Persona6 starts with Discord community agents, but the larger product is a reusable persona runtime. If that matches your wedge, we should talk.