yottoCode

The second major feature group comprises voice input and output, session resume, model switching, and real-time streaming. Voice notes sent to the bot are transcribed using Whisper, which runs locally on the Mac, ensuring privacy remains intact. Claude can respond with voice replies, creating a natural conversational flow. Session resume persists full conversation history to disk, allowing users to pause work and return later without losing context. Model switching lets users change between Haiku, Sonnet, and Opus mid-conversation, adapting cost and capability dynamically. Real-time streaming shows progressive message updates and tool progress indicators with smart rate-limited chunking, so users see Claude's thinking unfold. These capabilities transform Telegram into a full-fledged coding interface that matches desktop usage patterns.

The third feature group includes MCP Server integration, Files and Screenshots, and smart context. MCP servers allow users to connect, disconnect, and toggle external services without restarting yottoCode. This unlocks capabilities beyond coding—managing emails via G Suite, summarizing Slack conversations, creating Linear issues, or updating Notion databases. Files and screenshots enable Claude to send files directly to Telegram chat or request a screenshot of the Mac screen, providing visual feedback without physical access. Smart context works by replying to any message in Telegram to give Claude additional context; Claude acknowledges with emoji reactions, creating a natural Telegram-native conversation flow. These features elevate yottoCode from a coding assistant to a comprehensive productivity tool that integrates with existing workflows.

Overall workflow begins when a user sends a message to the bot on Telegram. This message travels through Telegram's cloud to the yottoCode bridge, which routes it to the Agent SDK. The SDK communicates with Claude, which performs operations on the local machine—reading files, editing code, executing commands—and streams responses back. The bridge orchestrates message handling, permission prompts, and voice/audio conversion. Security is maintained through outbound-only connections with no exposed ports, and authentication uses a Telegram user ID whitelist. Each bot instance manages precisely one project directory, but the multi-bot launcher runs multiple instances simultaneously for different projects. This architecture ensures that yottoCode runs on the user's machine, leveraging its full filesystem, tools, and persistent state without sandboxing or containerization.

Concrete use cases demonstrate yottoCode's practical value. A developer commuting by train receives an urgent bug report; they open Telegram, ask yottoCode to fix the authentication bug in login.ts, review the proposed edit via permission keyboard, approve, and see the build succeed—all from their phone. Another user connects MCP servers to manage their entire digital life: summarizing Slack channels, reading emails, creating calendar events, and updating Notion databases without switching apps. A DevOps engineer monitors deployment progress on their laptop while walking to a meeting; they receive streaming updates and approve changes using Telegram's inline keyboards. A manager uses voice commands to edit Google Docs hands-free, asking Claude to add summaries based on spreadsheet data. Each outcome delivers the same capabilities as sitting in front of a Mac, but accessible from any device with Telegram—watch, phone, tablet, laptop, or desktop.

Target users are developers with an active Claude Code subscription (Max, Team, or Enterprise) who need to maintain control over their development environment remotely. The platform currently runs on macOS 14.0+ with Apple Silicon (M1+), leveraging native features for privacy and performance. Pricing includes a free forever tier with one bot and 100 messages per month, and a Pro plan at $1.99 per month (or $2.98 per month billed yearly) with unlimited bots and messages. Alternatively, users can use an API key mode with live cost tracking per message and session. The product is not a chatbot or a cloud sandbox—it runs real code on the user's machine, providing a true desktop experience from Telegram. yottoCode delivers unmatched flexibility, allowing developers to ship code, debug, and manage projects from anywhere while maintaining full security and control.