Как экономить токены и управлять памятью AI агента в 2026 году Команда Convolife
The Convo Life team presents a tool for tracking AI agent token usage and context window status within Telegram bots. The system provides usage thresholds and recommendations for when to compact or reset sessions to avoid exhausting the context window. It can also be visually customized per bot.
Summary
The presentation introduces Convo Life, a system designed to help users monitor and manage the token consumption of AI agents, specifically by tracking how much of the context window has been used and how much remains in the current session.
The tool operates on a threshold-based alert system. When the context window is between 0–40% full, the situation is considered fine. From 40–60%, it is acceptable but warrants attention. At 60–75%, the system recommends running a 'compact' command to compress the session context. At 75–90%+ usage, launching a new chat session ('newchat' or 'newat') is strongly advised to prevent rapid exhaustion of the available token budget.
The system is implemented via a command sent to a Telegram bot. The presenter shares an example where the command was run and the context window was found to be 70% loaded, prompting an immediate recommendation to clear the session and run a checkpoint due to context overload. Implementation requires sending a specific prompt with usage examples through a skill or CLD interface, after which the bot understands the command.
The tool is also customizable in terms of presentation. Some bots display the context load as a plain percentage (e.g., '70%', '5%', '25%'), while others show a visual representation of how loaded the context window is. One example bot named 'Demniy' is shown responding with a detailed message: 'The text window is loaded at 0.2%, you still have 998,000 tokens left. Everything is fine.'
Key Insights
- The Convo Life team argues that context window usage above 75% warrants launching a new chat session ('newchat') to prevent rapid token depletion in the AI agent.
- The presenter demonstrates that a 70% context load triggers an automatic recommendation from the bot to clear the session and run a checkpoint, treating it as an overloaded state.
- The team explains that the tracking command is implemented by sending a specific prompt with usage examples through a skill or CLD interface, after which the AI agent interprets it autonomously.
- The presenter notes that the context window display is customizable per bot — some show a plain percentage while others render a visual indicator of context load.
- An example bot named 'Demniy' is shown responding with granular detail, reporting '0.2% loaded with 998,000 tokens remaining,' illustrating the system's capacity for precise token accounting.
Topics
Transcript
[0:00] The next team is Convo Life. Convo Life allows us to track the number of tokens we have already spent and those we have left to spend. That is, what size of context window do we have left in the current session ? So, there are several indicators. When we have a context window from 0 to 40%, then everything is fine. When it’s from 40 to 60, then, in principle, it’s normal. But you still need to pay attention and monitor. when 6075, then it is better to make a comp and the [0:30] compact command is our next command, looking ahead. Well, if we have 90+%, then it is better to launch newcha, or 75+, then it…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Игорь Зуевич
Сравнение Claude Code, Gemini и Codex — тесты на контент, анализ и изображения
The speaker is comparing AI coding assistants — Claude Code, Gemini, and Codex — by testing their ability to create image-generation agents. Each tool is being evaluated on how well it writes prompts and structures sub-agents for text-to-image and image-to-image tasks.
Экономия токенов и контекстного окна: настройка памяти нейросети – Compact сжатие контекстного окна
The transcript explains the 'Compact' command in AI agent sessions, which compresses the entire conversation history to reduce context window usage without switching sessions. It also covers related commands for full history clearing and automatic triggers for compact execution. The speaker recommends configuring automatic notifications or auto-execution when the context window reaches a certain threshold.
HR-агент на Claude Code: создаёт других сотрудников сам по запросу
The video demonstrates setting up a Telegram HR bot using Claude Code, walking through the process of creating a bot via BotFather, obtaining a Telegram user ID, and letting the code automatically generate the HR agent. The presenter shows how the system initializes itself by installing dependencies and launching the bot with minimal manual configuration.
Живые скиллы для Claude Code, Gemini и Codex | Как настроить super skills для ИИ агентов
The video explains how to create 'living' or 'super' skills for AI agents like Claude Code, Gemini, and Codex that self-update based on user feedback. The presenter demonstrates how these skills collect feedback after each task, store it in log files, and periodically incorporate that feedback into the skill's core instructions. This approach reduces token consumption and ensures skills continuously improve over time.
Планировщик для ИИ агента: 50 идей и готовые файлы для Claude, Codex и Gemini
The speaker describes their use of a scheduling/routing tool called Routins, which they compare to N8N, to automate daily delivery of curated YouTube videos on topics like Gemini and Claude. The system runs autonomously, collecting videos and generating concise summaries of their content without manual intervention.