NEW — Chatty now ships a full MCP server.Learn more →
PARTNERS — Earn 30% recurring commission.Join now →
Chatty - Open-source AI support bot for your website | Product Hunt

Trained on your content. Optimized for conversion.

An autonomous AI agent that captures leads, books meetings, and triggers actions. Zero coding, live on your site in five minutes.

Zero-Code, Full Control

Or drive it all through MCP

Captures Every Lead

Name, email, phone - automatically

Books Its Own Meetings

Straight onto your calendar

Start free 14-day trial

14-day trial · No credit card required

[ 01 / Transparent Fees ]

Pricing plans

Every cloud plan includes a full-featured 14-day free trial with no credit card required. The only free forever option is self-hosting our open-source stack on your own infrastructure.

14-Day Free Trial
[ PLAN: HOBBY ]

Hobby

$19/mo

billed monthly

Perfect for individuals, developers, and side projects.

  • 14-day free trial included
  • 1,000 weighted AI credits/mo
  • 10M training characters
  • 500 MB Media Storage (250 products)
  • 3 AI agents
  • Fast & Advanced AI models
  • AI Actions & Analytics
  • Guardrails & Notifications
  • Lead collection & API
  • Provider-managed AI routing
Start 14-day free trial
Popular choice14-Day Free Trial
[ PLAN: STANDARD ]

Standard

$99/mo

billed monthly

All in Hobby, plus advanced automation and multi-bot systems.

  • 14-day free trial included
  • 10,000 weighted AI credits/mo
  • 20M training characters
  • 2 GB Media Storage (1,500 products)
  • 6 AI agents
  • Daily Auto Train sync
  • Remove branding completely
  • Unlimited team members
Start 14-day free trial
14-Day Free Trial
[ PLAN: BUSINESS ]

Business

$399/mo

billed monthly

For enterprise scale, heavy traffic, and reseller options.

  • 14-day free trial included
  • 40,000 weighted AI credits/mo
  • 50M training characters
  • 10 GB Media Storage (Unlimited products)
  • Unlimited agent configurations (fair-use limits apply)
  • BYOK (Bring-Your-Own-Key) option
  • White-label configuration
  • Management Admin API
Start 14-day free trial
Only Free Option[ SELF-HOSTED / OPEN SOURCE ]

Self-Host on Your Own Server — $0 Forever

Want a 100% free setup? Self-host the open-source Chatty stack on your own infrastructure with Docker and PostgreSQL/pgvector, bring your own LLM API keys (OpenAI, Anthropic, Gemini, OpenRouter), and manage everything yourself.

Self-Host via GitHub

[ Taxes & Compliance ] All plans are subject to local tax system regulation.

[ 02 / Capabilities ]

Core features

A granular index of Chatty's feature set. Click on any block to see detailed configuration parameters and dashboard instructions.

[ 03 / Agent Control ]

Start with MCP

Chatty ships a full Model Context Protocol server. Point Claude, ChatGPT, or any MCP-compatible client at your account and run the entire dashboard - every bot, flow, campaign, and integration - from a conversation instead of clicking through screens.

Connect your agent

OAuth 2.0 + PKCE

RFC 7591/8414-compliant dynamic client registration - no shared secrets pasted into a config file.

53 real tools

Bots, flows, campaigns, knowledge, inbox, leads, calendar, guardrails, team, billing, GDPR export.

Same data, same rules

Every tool reads and writes the exact tables the dashboard does - nothing simulated, nothing mocked.

Scoped access

read / write / knowledge / actions / admin scopes, so an agent only gets what it needs.

Recommended

Install the Chatty Codex integration

The public plugin lets any Codex user add Chatty from GitHub. It ships the Chatty icon, starter prompts, skill guidance, and the hosted MCP endpoint in one integration.

  1. Open Codex, then go to Settings, Plugins, Add, and choose Add plugin marketplace.
  2. Paste the public Chatty repository values shown here, then add the marketplace.
  3. Install the Chatty integration, keep it enabled, and approve OAuth on first use.
  4. Ask Codex to audit bots, update knowledge, triage inboxes, manage campaigns, or create booking-ready assistants.
Open dashboard

Connection value

Source
https://github.com/PersonaliAI/chatty.git

Git ref
main

Sparse paths
.agents/plugins/marketplace.json
plugins/chatty-integration

The client opens a standard OAuth 2.0 authorization flow on first connect - approve it once, no API key to copy anywhere.

[ 04 / Help Center ]

Build a bot people trust

Practical operating guides for a useful, safe, and conversion-ready Chatty deployment. Each guide is written for the person responsible for the customer experience - not just the installation.

Browse documentation
01Getting started

Launch your first Chatty bot

Start with one focused use case: answer product questions, qualify visitors, or route support requests. Give the bot a clear identity, train it from trusted sources, then test it like a real visitor before publishing.

OutcomeA trained, branded assistant live on a production page with a tested fallback path.

Create the bot

Open the dashboard, create a bot, and set the name, welcome message, tone, language behavior, and first suggested prompts. Keep the first welcome short and action-oriented so visitors know exactly what to ask.

Train the source of truth

Add your website, help docs, pricing page, policy pages, PDFs, or pasted text. Wait for indexing to finish, then ask the bot ten questions your customers actually ask.

Publish safely

Copy the embed script from the Embed tab and place it before the closing body tag. Add your production domain to the allow list before sending traffic to the widget.

Before you ship

  • Bot identity and welcome message are set
  • At least one authoritative knowledge source is indexed
  • Embed script works on a staging or production page
  • Fallback contact or handoff route is visible
02Knowledge

Keep answers accurate with a healthy knowledge base

A strong bot is only as good as the material it can cite. Treat the knowledge base like a living help center: structured, current, and biased toward canonical sources instead of broad crawls.

OutcomeReliable answers that stay aligned with your current product, pricing, and policies.

Use canonical sources

Train from pages that contain final policy, pricing, setup, billing, delivery, return, and troubleshooting information. Avoid duplicate drafts and outdated campaign pages.

Close missing-answer gaps

Review unanswered questions and low-confidence conversations weekly. Add a short article or source snippet for every recurring gap instead of relying on prompt changes alone.

Automate freshness

Enable Auto Train for sources that change often. Daily sync is best for pricing, inventory, status, and policy pages; weekly or monthly works for slower documentation.

Before you ship

  • Pricing and policy pages are indexed
  • Old or duplicate sources are removed
  • Unanswered questions are reviewed regularly
  • Auto Train is enabled for changing content
03Safety

Secure a bot before publishing

Security is part of the customer experience. Lock the widget to approved domains, keep integration keys scoped, and define exactly when Chatty should stop and hand the conversation to your team.

OutcomeA public assistant that answers confidently without leaking access or pretending to be a human.

Restrict where it runs

Add every approved website to the Allow List before sharing the embed code. This prevents copied scripts from being used on unknown domains.

Limit integration access

Use separate API keys for separate systems and grant only the scopes each workflow needs. Rotate keys when a teammate or vendor no longer needs access.

Design the handoff

Add clear escalation rules for billing disputes, account access, urgent issues, sensitive personal data, and questions the bot cannot answer with confidence.

Before you ship

  • Production domains are allow-listed
  • API keys are scoped by integration
  • Human handoff rules are written
  • Abuse and off-topic guardrails are enabled
04Conversions

Turn useful conversations into qualified leads

Lead capture works best when it feels like the natural next step, not a form dropped in front of the visitor. Ask for contact details only when the conversation has enough intent to justify it.

OutcomeMore qualified conversations reaching your inbox, CRM, calendar, or sales team with context intact.

Ask at the right moment

Trigger lead collection after pricing, demo, custom quote, support escalation, or availability questions. Keep the form short: name, email, and phone only when phone follow-up is useful.

Preserve conversation context

Send the transcript, visitor page, lead fields, and detected intent to your team so follow-up starts from the conversation instead of a blank record.

Book when intent is high

Use calendar booking for demo requests, consultations, onboarding calls, and urgent support. Let the visitor choose a real available slot before they leave the site.

Before you ship

  • Lead trigger is tied to intent
  • Required fields are minimal
  • Email or webhook delivery is tested
  • Booking is connected for high-intent flows
05Operations

Run the inbox and human handoff well

The inbox is where automation meets real customers. Use ownership, status, and feedback consistently so the bot improves instead of burying work.

OutcomeA clean support rhythm where AI handles repeat questions and humans own the conversations that need judgment.

Own active conversations

Assign conversations deliberately and keep team availability current. When a teammate takes over, pause AI replies so the visitor does not receive conflicting answers.

Use statuses consistently

Separate open, pending, resolved, and follow-up conversations. This makes response time, missed handoffs, and unresolved issues visible.

Turn corrections into training

When a human fixes an answer, capture that correction as a knowledge improvement. Repeated inbox fixes should become articles, sources, or guardrail changes.

Before you ship

  • Conversation ownership is assigned
  • AI pause is used during human takeover
  • Resolved status is applied after follow-up
  • Corrections feed the knowledge base
06Codex plugin

Add Chatty to Codex with the plugin

Use the public Chatty plugin when you want the easiest Codex setup. The plugin contains the MCP endpoint and Chatty instructions, so users only add the marketplace source once and approve OAuth when Codex connects.

OutcomeCodex can manage Chatty bots, knowledge, inboxes, campaigns, bookings, and analytics through the hosted MCP server.

Add the marketplace

In Codex, open Settings, Plugins, Add, then Add plugin marketplace. Use source https://github.com/PersonaliAI/chatty.git, git ref main, and sparse paths .agents/plugins/marketplace.json and plugins/chatty-integration.

Install Chatty

After the marketplace loads, install the Chatty integration. Codex will show it as a normal plugin with the Chatty icon, description, starter prompts, and MCP connection.

Connect your account

Ask Codex to use Chatty. The first run opens OAuth; approve only the scopes you want, then Codex can audit bots, update articles, triage inboxes, and manage booking flows.

Before you ship

  • Marketplace source is the public GitHub repo
  • Sparse path is plugins/chatty-integration
  • Chatty plugin is installed and enabled
  • OAuth approval completes on first use
[ 05 / Common Inquiries ]

Questions

Everything you need to know about Chatty's training mechanics, costs, safety layers, and white-label setups.

Ready to convert more visitors?

Zero coding. Live on your site in five minutes. Cancel any time during your trial.

Start free 14-day trial