core
Production-grade models
A curated stack of conversational models tuned for persona consistency. Stable inference, low latency, and safety controls baked in at the model layer — ready to ship without adding your own ops layer on top.
01 AI assistant builder · multi-channel
ChameleAI is a platform for building, testing, and shipping AI assistants. Author the voice, tune the behavior, deploy in minutes — Telegram today, Discord and WhatsApp next, the rest on the roadmap. No prompt files, no server config, no LLM ops.
25,000+ Telegram bots deployed on the underlying platform since 2023
02 The pipeline
Compose the persona in a structured editor — voice, knowledge, scene, refusal lines, system prompt. No freeform prompt engineering required.
Chat with your bot in a browser sandbox. Iterate on tone, catch edge cases, validate behavior before users do.
Push to Telegram with one click. Manage instances, rotate models, and roll back from a single dashboard.
03 What's in the box
core
A curated stack of conversational models tuned for persona consistency. Stable inference, low latency, and safety controls baked in at the model layer — ready to ship without adding your own ops layer on top.
authoring
Forms, not freeform prompts. Define voice, knowledge, scene, and tone in a structured editor.
deploy
First-class deployment pipeline. Your bot ships as a real @username, stable and observable.
models
Free tier ships on Qwen 3.6 27B tuned in-house. Higher tiers unlock Claude 4.7, GPT-5, Gemini 2.5 Pro, and DeepSeek V3.5 via OpenRouter — picked per assistant, swappable any time.
04 Built for
Build expressive Telegram bots without managing infrastructure or wrangling prompt files. Spin up a persona, test it, ship it. The rest of the platform — moderation, deployment, billing — is already wired.
Start a free workspaceSkip the LLM ops layer. Ship persistent AI agents on Telegram in an afternoon. Bring your own models in higher tiers, or run on the curated stack with one less thing to maintain.
Open the dashboard05 Where it shows up
support
Answer docs-level questions before they hit your on-call. The assistant quotes the changelog, links to the right runbook, and triages repro reports into the structure your engineers already expect — in the channel your users already live in.
internal
Onboarding helpers, IT/HR FAQ, ops runbook lookups. The assistant lives in the team's existing chat, with role-aware access to the right knowledge base.
creators
Keep engagement warm between drops. The assistant handles recurring member questions in-voice and reposts context for new subscribers.
local
Quote daily specials, take reservation requests, answer hours
and location queries — in the local language, in your brand
voice, at the hours your staff can't cover. Stable
@username customers can trust.
06 Pricing
For solo creators getting started.
$15 / month
or $150 / year
For active creators shipping multiple bots.
$35 / month
or $245 / year — save $175
For teams and serious operators.
$75 / month
or $500 / year — save $400
Free tier (Qwen 3.6 27B) is unmetered. Frontier-model usage is metered per plan; add-on packs start at $2.99 and roll over for 12 months. Cancel anytime, billed in USD.
07 Questions
Conversational AI personas deployed as Telegram bots — assistants, branded characters, customer-service agents, narrative experiences, and interactive bots for creators. You define the voice and behavior; we handle the model, moderation, and deployment.
Connect your bot token from BotFather, choose a persona, click deploy. ChameleAI handles message routing, model inference, rate-limiting, and observability. Roll back or swap models from the dashboard at any time.
A curated stack of conversational models tuned for persona consistency, with safety classifiers and content guardrails on every request. Pro includes the standard tier; Limitless and Studio unlock the full model catalog including higher-context variants.
Anything that fits the Acceptable Use Policy — assistants, characters, helpers, coaches, narrative experiences. Safety controls are enforced at the model layer so you can ship without building your own moderation stack.
Yes. Cancel from your dashboard. Your subscription stays active until the end of the current billing period. No retention friction, no email lock-in.
The free tier ships on Qwen 3.6 27B, tuned in-house, with no per-message limit. Frontier models — Claude 4.7, GPT-5, Gemini 2.5 Pro, and others — run on a metered allowance bundled with each paid plan. If you need more, add-on packs start at $2.99 and roll over for 12 months.
08 Begin
Free to start. Upgrade when you ship.