Everything SMS-24 can do for your AI workforce

Eleven feature pillars covering messaging, voice, Android automation, network routing, and phone-reputation antifraud — all multi-tenant, all BYOK.

AI personas

A long-lived AI identity with name, role, biography, free-form system prompt, and uploadable knowledge base (PDF, DOCX, TXT, Markdown). Choose OpenAI, Anthropic Claude, or Google Gemini per persona. Link to N Telegram + WhatsApp accounts (many-to-many).

Messaging auto-reply (Telegram + WhatsApp)

Inbound messages arrive on the messenger-hub Redis Stream, get joined with the persona context (system prompt + recent conversation history), and a reply is generated via your BYOK API key. Telegram via Telethon (full MTProto session); WhatsApp via Baileys multi-device bridge.

Voice agents for Telegram calls

Four Pipecat-powered pipelines:

• cascade — Deepgram STT + OpenAI/Anthropic/Gemini LLM + Cartesia TTS. ~$0.06 / minute.
• realtime — OpenAI Realtime API over a single WebSocket. Lowest latency.
• realtime_gemini — Google Gemini Live, multimodal.
• cascade_local — self-hosted Whisper STT + MMS-TTS-RUS. Zero per-minute provider cost.

Persona function-calling (LLM tools)

Link Android device-action templates to a persona and the LLM sees them as OpenAI-compatible tools. Mid-conversation the model can call send_whatsapp_text(phone, body), send_telegram_text(username, body), install_apk_url(url), or any custom adb_shell / uiautomator_script action.

Emulated Android device automation

Each user gets redroid Android emulator instances managed by the operator. Devices share the netns of a per-device ProtonVPN sidecar — every byte egresses through the chosen country. ADB control, uiautomator2 declarative scripts, google_sign_in via Chrome automation, cron scheduler, live screen control via scrcpy + WebCodecs.

Per-user API key vault (BYOK)

Bring your own keys for six providers — OpenAI, Anthropic, Google Gemini, Deepgram, Cartesia, ElevenLabs. Fernet-encrypted at rest, decrypted only inside the worker that calls the provider. Every LLM, STT, and TTS call bills the persona owner's account.

ProtonVPN orchestrator

Multi-account orchestrator with per-country sessions. Each session runs an OpenVPN container exposing SOCKS5 (port 1080) and HTTP (port 3128) on the internal docker network. Operator uploads .ovpn configs per country in the admin UI; sessions auto-register into the shared proxy pool consumed by the scraper, OSINT, messenger-hub, and Android devices.

Smart proxy mode for scraping

Three-state policy per scrape target: direct (never paid proxy), auto (try direct first, fall back to Decodo on HTTP block signals), always (always paid pool). Per-host telemetry shows the direct-vs-proxy ratio.

Real-time phone-reputation antifraud

One GET /api/v1/lookup call returns a verdict in ~50 ms. Twelve orthogonal risk signals — public-board appearances, OTP collisions across services, cross-country indices, Telegram presence, carrier line type, fraud-ring cluster membership, and more.

Multi-tenant data isolation

Every user-scoped entity carries owner_user_id UUID NOT NULL. Filters live at the SQL level; cross-tenant reads return HTTP 404 (not 403, so existence is never telegraphed). Secrets Fernet-encrypted at rest.

API surface

Versioned REST under /api/v1/*, GraphQL via Strawberry, WebSocket for live event streams, MCP server at /v1/mcp/server.

See use cases → · See pricing → · Start free trial →

---

Related

• Use cases
• Business
• Pricing
• Check
• Persona Function Calling Llm Tools
• Voice Agents Telegram Calls Pipecat