Teleton Agents

📌 Teleton Agent v0.11.3 A few minors changes since 0.11.1 💻WebUI Refresh The dashboard now shows live agent activity, Telegram identity and token usage for the day, week or last 30 days and more visual changes. 💬Inline Rich Message Buttons Agents can add URL and copy buttons directly inside native Telegram Rich Message. 💫New model options GPT-6 Sol, Luna and Astra; Claude Opus 5.5 and Fable 5.1; Gemini 3.8 Flash; Grok 4.7. You can also use custom OpenRouter model IDs, with reported usage costs preserved. 🏷More reliable agent Improved RAG retrieval and local model selection. Fixes prevent duplicate messages and task execution, make config and core memory writes atomic, and ignore unknown commands in groups. 🔧Maintenance Updated dependencies and build runtime, with streamlined CI and Docker release builds. Update your agents: npm install -g teleton@latest ➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 👤 @zkproof

🤖Note about Cocoon: the Cocoon Network currently has only one active proxy and unfortunately 0 workers, which means that no AI inference is available at the moment on Cocoon Network. (it’s been a few days) As a result, Gocoon and the Cocoon provider in Teleton Agent will not work until the public network is restored. Since there is no TON Foundation and no contact at Telegram working in the interest of the TON community, it remains unknown whether the Cocoon network will have available workers in the future. (or not)

📌 Teleton Agent v0.11.1 What’s new since v0.9.0? 💬 Native Rich Messages

Agents can now send, read and edit rich Telegram messages with formatted content, media and interactive buttons.

🤖 New Provider: Grok Build

Keyless authentication through grok login, with Grok 4.6 by default. (Included in Twitter Premium+)

✨ 50+ New Model Options

GPT-5.6 Sol/Terra/Luna, Claude 5, Gemini 3.6/3.5, Grok 4.6/4.5, Kimi K3, GLM 5.2, MiniMax M3, DeepSeek V4 and more.

🔎 Global Telegram Search

Search messages across your account and posts across public channels

🧭 Smarter Tool Discovery

Faster, more accurate tool selection designed for large tool catalogs

🧠 Faster RAG

Memory indexing and vector search now run in the background without blocking Telegram.

🧩 Plugin SDK 2.1

New TON, DEX, DNS, Highload Wallet, Telegram and collectible capabilities.

npm install -g teleton@latest

💎 Teleton Agent v0.9.0 What's new since v0.8.6? 🤖 New Provider : Gocoon - Pay-per-use over a TON payment channel, no API key - Set up / top-up / withdraw your channel from the CLI or WebUI - Replaces the cocoon provider 📟Full WebUI Redesign Complete redesign of the WebUI, + new Cocoon page, and moving the logs from the dashboard to the new Logs page. 🔎Smart Tool Discovery: - ToolSearch ON by default: the agent loads tools on demand via a tool_search tool - Sharper tool selection (rewritten descriptions, query enrichment, balance/holdings boost) 🛡Tool Access Control - Per-tool access levels: All / Allow List / Admin / Off - New Access Policy dashboard + Allow Lists (Users / Groups / Admins) - Per-tool user / bot mode enforced ✨ New Models - Claude Opus 4.7, GPT-5.5, Gemini 3.1, Grok 4.3 - Haiku 4.5 is the new default of Anthropic - Codex CLI as a keyless auto-detect provider - $IN / $OUT pricing shown everywhere And lots more: get_available_gifts tool, full SDK docs site, npm provenance + CI hardening, 9 security fixes, dependency bumps, etc. See changelog

⚠️ Upgrading from 0.8.x: cocoon -> gocoon and claude-code provider has been removed, auto-migrated on load (claude-code users need an Anthropic key)

Update your agents: npm install -g teleton@latest ➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 👤 @zkproof

🤖 Gocoon - Cocoon Client in Go I built a Cocoon client in Go. It follows the same wire protocol and contracts interface as the existing C++ client. It’s cross-platform, with a simple standalone CLI. Connect to Cocoon Network, build LLM-powered mini-apps, bots, or AI agents with Gocoon. Gocoon on GitHub: https://github.com/TONresistor/gocoon

📌 Teleton Agent v0.8.6 What's new since v0.8.0? Dual-Mode: Bot + User

- Run as a Telegram Bot (Grammy) or GramJS Userbot - Switch mode at runtime from the WebUI - Bot mode: native message streaming, inline buttons, callback handlers - 67+ bot tools

HTTPS Management API

- Control your agent remotely over HTTPS with API key auth. - Manage your agent :Start, stop, restart, status -`teleton start --api`

Hook System v1

- User-configurable keyword blocklists with NFKC normalization - Context triggers injected into LLM prompt on keyword match

Heartbeat Autonomy

- Agent wakes on a timer (default 60min) - Reads HEARTBEAT.md from workspace, acts on tasks or stays silent.

TON Proxy Module

- direct p2p connection to TON Network - health checks, auto-restart on crash etc. - Live toggle from WebUI

QR Code Telegram Login

- Scan a QR from the setup wizard instead of phone+code

5 New LLM Models

- GPT-5.4, GPT-5.4 Pro Gemini 3.1 Pro, Gemini 3.1 Flash / Lite.

And lots of improvements, like a redesign of the webUI, new SDK methods for Hooks, etc. Update your agents: npm install -g teleton@latest ➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬 @zkproof

🌐BOARDS.TON To test x402 payments in a real-world environment, I built a forum: boards.ton, designed for AI agents. Posts and replies are protected via HTTP x402 (TON) and require small micropayments (0.01–0.05 TON). Any agent can participate, you don’t need a Teleton agent specifically. You can also use OpenClaw or even Claude Code. 🧠 Skills for agents: boards.ton/SKILL.md 🤖 For Teleton agents: You can either use the skill in YOLO mode or install the “boards” plugin in the WebUI. ➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬@zkproof

X402 TON & Jettons SDK v1 TypeScript SDK implementing the x402 open payment standard for TON Sign, verify, and settle HTTP 402 payments with native TON and USDT. Built for API services, AI agents economy and autonomous commerce. You can also deploy your own payment facilitator server and verify payments for third parties As demo i deployed a server that you can use to verify x402 payments : x402.resistance.dog

An interesting thing to do would be a decentralized network of facilitators on TON; each facilitator could become a node and contribute to verifying HTTP 402 payments for fees.

https://github.com/TONresistor/x402-ton

📝Guide: Deploy your Teleton Agent Telegraph: https://telegra.ph/Deploy-Your-Teleton-Agent-03-03 X Article: https://x.com/teletonagent/status/2028709246842417376 GitHub: https://github.com/TONresistor/teleton-agent/blob/main/GETTING_STARTED.md Join the community: @TeletonAgentHQ

📌 Teleton Agent v0.8.0 What’s new since v0.7.5 ? Inline Mode for Plugins Plugins can now register as inline bots with colored buttons (green, red, blue), callback handlers with pattern matching, and rate limiting. 3 New LLM Providers & free Model: - ZAI • GLM4.7 flash (free!) - HuggingFace (8 models via 1 token) - MiniMax M2.5 GramJS Layer 223 Upgraded to the latest Telegram protocol. The agent now sees participant ranks and titles. admins, moderators, and custom roles are now displayed in conversation context. 29 New Plugin SDK Methods DEX trading, DNS auctions & bidding, TON Site ADNL records, jetton analytics, scheduled messages, chat history & dialogs, Stars transactions, gift collectibles & NFT transfers. Fixes & changes:

- Moonshot refactored from hardcoded models to native pi-ai integration. - Transcript files and secrets directories now created with restricted permissions (0o600/0o700). - API keys properly masked in the web dashboard. - SDK Documentation update - Silent error catches replaced with structured debug logging across jetton price and workspace tools.

Update your agents: npm install -g teleton@latest ➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬@zkproof

📌 Teleton Agent v0.7.5 - News: YOLO Mode (Coding Agent)

The agent can now execute system commands, install packages, manage services, and check server status directly from Telegram. Perfect for managing a VPS. Disabled by default for safety, you choose when to turn it on. Every command is logged for full transparency.

GramJS Layer 222 Fork

Upgraded Telegram connectivity to the latest protocol version. This brings support for new Telegram features and removes the need for manual patches after each update.

Send, accept & decline Gifts offer The agent can now handle Telegram gift offers in real time : send, accept, or decline them automatically when they come in. Admin-Only Access Policy New security option that restricts the agent to only respond to designated admins. Now the default for fresh installations. DNS Set-Site dns_set_site links a .ton domain to a TON Site via ADNL address for decentralized website hosting. TON Balance (Telegram intern wallet) The agent can now check your Telegram internal TON wallet balance, not just Stars. Live Token Usage Tracking The web dashboard now shows how many AI tokens you're consuming in real time, so you can monitor costs as they happen. Toncenter API Key Optional API key support for faster and more reliable TON blockchain queries. Session Auto-Pruning Sessions older than 30 days are automatically cleaned up at startup. - Changes:

Smarter tool selection is now on by default. Instead of sending all 125+ tools to the AI on every message, the agent picks only the most relevant ones. Faster responses, lower costs. Better tool descriptions across the board for more accurate automatic tool selection. New installations default to admin-only access instead of open = more secure out of the box. Setup wizard now includes a Coding Agent option and clearer policy choices. Web dashboard redesigned with easier provider switching, better tool and plugin management. Gift catalog improved with search, sorting, pagination, and resale filters. Gift resale error messages are now human-readable instead of cryptic error codes. Ston.fi DEX integration upgraded to the latest SDK with better transaction safety. TON addresses and identifiers no longer break when displayed in messages (formatting fix).

- Fixes:

The typing indicator now stays visible the entire time the agent is thinking, instead of disappearing mid-response. Login flow fixed for accounts that require payment verification on Telegram. Gift sending now works correctly for all user types. Docker image build fixed after the previous update removed legacy files.

Update your agents:

npm install -g teleton@latest

➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬@zkproof

📌 Quick update - v0.7.3 + 0.7.4 + claude code provider w/ auto detection and refresh of the access token of your claude code config + support Telegram Anon (+888) number + inject reply-to message context in prompt + 7 new Telegram tools

manage scheduled messages, voice transcription, collectibles infos, set username channel...

+ auto-transcription of voice/audio messages + add more available models

new: claude-sonnet-4-6, gpt-5/5-pro/5-mini/5.1, o4-mini, codex-mini, gemini-3-pro/flash, gemini-2.5-flash-lite, grok-4-1-fast, grok-code, deepseek-r1-0528/v3.2, qwen3-coder/max, nemotron-nano, sonar-pro...

+ WebUI config and dashboard refactor with pill-bar tabs + memory browser + image preview + more options + lot of UX improvements

npm install -g teleton@latest

➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬@zkproof

🤖

📌 Quick update: Setup is easier now; with the latest release, the web UI is directly accessible from setup to execution of your agent with the command : teleton setup —ui ➖➖➖➖➖➖➖➖➖➖➖➖➖ For those who want to build services on top of Teleton Agent, I recommend waiting for a stable v1.0 before any production/mass distribution. 🙏 We look forward to your feedback and bug reports in the @TeletonAgentHQ group. 🤓 ➖➖➖➖➖➖➖➖➖➖➖➖➖ 💬 @TeletonAgentHQ | 💬@zkproof

The Teleton website has been updated, and the documentation published at docs.teletonagent.dev website: teletonagent.dev ➖➖➖➖➖➖➖➖➖➖➖➖➖ 📢 @TeletonAgents | @TeletonAgentHQ

🤖Teleton v0.6.0 What’s new today ? MCP Support MCP client integration with dedicated WebUI tab and CLI commands (teleton mcp add/remove/list). Plugin Marketplace - Browse, install, update and uninstall community plugins from the WebUI. - Per-plugin encrypted secrets management. - Bulk enable/disable and scope control per plugin (All, DM, Group, Admin). Web Search (Tavily) New web_fetch and web_search tools. Free tier: 1,000 req/month. Built for AI agents.

- Get a free Tevily API key at app.tavily.com (30 second, no credit card) and get a free plan 1000 req/ month

Tool RAG Semantic tool selection: only the most relevant tools are sent to the LLM per message. Enable/disable from the Dashboard. WebUI Improvements - New pages: Config, MCP, Dashboard settings. - Several improvements and more configuration options added Under the hood All dependencies upgraded, plugin SDK guide, more docs, token optimizations etc.

  npm i -g teleton@latest

➖➖➖➖➖➖➖➖➖➖➖➖➖ 📢 @TeletonAgents | @TeletonAgentHQ

Teleton Agent v0.5.1 WebUI Dashboard - Real-time logs (SSE), tools config, plugins, soul editor, memory search, workspace manager - Bearer token auth, localhost only Plugin SDK (@teleton-agent/sdk) - First public release on npm: 53 typed methods - Full TypeScript autocompletion for plugin developers - npm install -D @teleton-agent/sdk

npm install -g teleton@latest

➖➖➖➖➖➖➖➖➖➖➖➖➖ 📢 @TeletonAgents | @TeletonAgentHQ

I’m participating in TF’s initiative for Ai Agents project builders on TON, it’s a channel where we will debate and share on AI agents 🤖 Join the channel : https://t.me/ai_dev_wall