Flagship
Five AI providers, one platform
Claude, GPT, Gemini, Llama (via OpenRouter) and Qwen, pre-wired and ready. Setting up an AI agent takes a couple of minutes.
Product capabilities
Everything in one platform. Nothing locked behind seat tiers.
Multi-model AI, multi-widget, WebSocket-native messaging, built-in analytics, HMAC webhooks and pentest-verified multi-tenancy: a flat $19.99/month, not per seat.
30 capabilities · 7 product areas · 5 AI providers, BYOK · 14-day trial on paid plans
01 · AI agents
Multi-model AI agents with BYOK
NexTalk lets you pick the AI model behind every widget. Connect Anthropic Claude, OpenAI GPT, Google Gemini, Meta Llama or Alibaba Qwen. Bring your own API key, pay providers directly at cost, and switch models in seconds without lock-in or per-resolution surcharges.
BYOK: bring your own API key
Plug your existing Anthropic, OpenAI, Google or OpenRouter key. Keys are AES-256 encrypted at rest and only decrypted inside the AI worker for each request. You pay the LLM provider directly, with no per-resolution markup from NexTalk.
AI scheduling per widget
Define when AI handles chats and when they go to your humans. Off-hours, weekends, holidays, lunch breaks: your call. Each widget has its own schedule and timezone.
Escalation rules to humans
Configure keyword triggers and an “AI can't answer” fallback. Full conversation context transfers to the human queue, so your agent picks up where the AI left off.
Custom system prompts per widget
Each widget has its own AI persona, tone, knowledge boundary and brand voice. Test prompts before going live.
02 · Real-time messaging
WebSocket-native live chat infrastructure
Built on a native WebSocket gateway backed by Redis pub/sub for cross-instance fan-out. No long-polling, no third-party realtime SDKs, no surprise latency spikes at scale.
Flagship
Typing indicators and read receipts
Customers and agents see who's typing in real time. Read receipts confirm delivery, building trust on both sides of the conversation.
Visitor presence and page tracking
See which page the visitor is on right now, how long they've been browsing, and which pages they viewed before chatting: full context before you say hello.
Browser and sound notifications
Native browser push for new chats, configurable sounds per event, and Do-Not-Disturb windows. Agents never miss a conversation.
Auto-reconnect with exponential backoff
Capped at 30s, no message loss. Works behind corporate proxies, VPNs and flaky mobile networks.
03 · Multi-widget
One workspace, many widgets
Run different chat experiences across your sites, products, landing pages or apps, all from a single NexTalk workspace. Each widget gets its own branding, AI agent, schedule, domain restriction and analytics.
Flagship
Independent widget configs
Brand colors, greeting text, position (bottom-left or right), agent avatar, welcome card: each widget configured separately. No shared “master” theme to fight.
Domain allowlist per widget
Lock a widget to one or more domains. Any embed attempt from elsewhere is rejected, which prevents widget theft and unauthorized use on other sites.
One-line embed (Shadow DOM)
Drop a single
<script>tag. The widget renders inside a Shadow DOM with z-index 2147483647, so it never conflicts with your CSS, even with aggressive Tailwind resets or legacy Bootstrap.Multi-language ready
Per-widget locale, RTL support and translation hooks. Greeting messages, error states and offline notices are translatable strings, not hardcoded English.
04 · Analytics and reports
Visibility into every conversation
Built-in analytics dashboard plus exportable reports, so you don't need to bolt on Mixpanel or Amplitude just to know how your support is doing. KPIs refresh in real time over WebSocket.
Flagship
Live KPI dashboard
Active conversations, agents online, queue depth, average response time, AI-handled vs human-handled split, updated live over the dashboard WebSocket channel.
Hourly volume heatmaps
Heatmap of chat volume by hour and weekday. Spot patterns at a glance: weekend spikes, lunch lulls, the 9am Monday tidal wave.
Satisfaction surveys (CSAT)
Post-chat star rating with optional written feedback. CSAT trend lines and per-agent breakdowns in the reports view.
CSV export and API access
Download conversation logs, agent performance and CSAT data as CSV for your BI tool. REST API for programmatic access on the Enterprise plan.
05 · Team management
Built for support teams of any size
Roles, departments, invites and presence: everything you need to run a real support operation, not a one-person side project. Per-tenant isolation keeps data safe even at multi-brand agencies.
Flagship
Roles: agent, admin, super-admin
Granular permissions per role. Agents handle chats. Admins manage settings, billing and team. Super-admins run multi-tenant deployments.
Departments and routing
Create departments (Sales, Tech Support, Billing). Route chats by widget or page context. Transfer between departments mid-conversation.
Canned responses with shortcuts
Save common replies as templates. Agents trigger them with
/shortcutsyntax, like/refund,/welcomeor/escalate. Reduce handle time without sounding scripted.Desktop app for agents (Linux, Windows, macOS soon)
Native desktop client with system tray, deep-linking from notifications to chats, and presence sync. Built on Electron, but lean.
06 · Integrations and webhooks
Plays nicely with the rest of your stack
Outgoing webhooks with HMAC-signed payloads, REST API on Enterprise, and Stripe billing. No locked-down integrations marketplace that charges per connector.
Flagship
Outgoing webhooks (HMAC-signed)
Subscribe any URL to events: new chat, new message, conversation closed, CSAT submitted. HMAC-SHA256 signatures let your endpoint verify authenticity. Works with Slack, Discord, Zapier, Make, n8n and custom systems.
Slack and Discord notifications
Ready-made templates for Slack and Discord. New high-priority chats ping one channel; 5-star CSAT ratings land in another. Two-click setup.
REST API access (Enterprise)
Programmatic access to conversations, messages, agents, widgets and analytics. OpenAPI 3.1 spec, JWT-authenticated, rate-limited.
Stripe billing built in
Subscription, trial, prorated upgrades and dunning, all handled by Stripe. No PCI scope on your end, no homemade billing bugs.
07 · Security and compliance
Enterprise security by default
Penetration-tested in March 2026 by an independent team, with the findings remediated. Progressive rate limiting, strict CSP, JWT in HttpOnly cookies and AES-encrypted secrets.
Flagship
JWT in HttpOnly cookies
Access and refresh tokens live in HttpOnly, Secure, SameSite cookies, not in localStorage, which protects them from XSS token theft. Refresh tokens are rotated per use and stored hashed in the database.
AES-256 encrypted API keys
Your AI provider keys are encrypted at rest with per-tenant keys and decrypted only inside the worker process for each LLM call. Never returned to the dashboard, never logged.
Progressive rate limiting
Per-IP, per-tenant and per-endpoint rate limits with exponential cooldown. Brute-force, credential stuffing and scraping are throttled automatically.
Strict CSP and security headers
Content-Security-Policy, HSTS, X-Frame-Options, Permissions-Policy and Referrer-Policy, all set sensibly. The widget injects only into its Shadow DOM, never into the host page's global CSS.
Tenant isolation, pentest-verified
Every database query is filtered by the tenant_id from the JWT. Multi-tenant isolation was verified in the pentest: no horizontal privilege escalation across tenants.
Ready when you are
Ship a better support experience this week.
$19.99/month flat. 14-day free trial. No credit card required. Setup takes under 2 minutes: drop one script tag and you're live.
No credit card · Cancel anytime
- 2-minute
setup - Your AI,
your key - Human
handoff - Real-time
team inbox