Telegram Dev

Keep Telegram bots and AI coding agents on the current Bot API.

Telegram changes the Bot API every month or two, and both code and models fall behind. Telegram Dev has three parts:

telegram-dev check reads a Python bot (aiogram, python-telegram-bot, pyTelegramBotAPI or raw HTTP calls) and lists Bot API calls with retired parameters or invalid values. It runs in a terminal, in CI as a GitHub Action, or as an agent tool.
Bot API tools for agents: method and type definitions, a request checker and the changelog by version, built from the official pages.
A documentation server and skill for Claude Code, Codex and Cursor: search over the official Telegram and TON docs (Bot API, Mini Apps, MTProto, TDLib, bot libraries, TON Connect, Tolk), refreshed automatically.

No Telegram account, bot token or API key is needed: everything reads public documentation.

Check a bot project

telegram-dev check checks every Bot API call in a Python bot — aiogram, python-telegram-bot, pyTelegramBotAPI or raw HTTP requests to api.telegram.org — in one run, and prints only the findings:

uvx telegram-dev check ./my-bot

$ uvx telegram-dev check ./tutorial        # a popular open-source aiogram 3 tutorial
code/03_buttons/bot.py:79: warning: types.KeyboardButton: `request_user` was retired in Bot API 7.0: Renamed the field request_user in the class KeyboardButton to request_users. The old name will still work for backward compatibility. ...
code/08_inline_mode/handlers/inline_mode.py:48: warning: inline_query.answer: `switch_pm_text` was retired in Bot API 6.7: ... replacing the parameters switch_pm_text and switch_pm_parameter of the method answerInlineQuery with the parameter button of type InlineQueryResultsButton. ...
Bot API 10.4; 95 files with bot code; libraries: aiogram ==3.14.0 ==3.7.0 >=3.24.0 ...; 0 errors, 5 warnings

A wrong value is an error, for example InlineKeyboardButton(text="Delete", style="red") gives style: must be one of danger, success, primary.

Errors are values outside the documented list and unknown methods or parameters in raw API requests. Warnings are parameters Telegram retired (with the version and the changelog entry that replaced them), keyword arguments that are neither Bot API parameters nor known library options, and bot methods with no Bot API counterpart. It reads code without running it, so it does not follow variables or check required parameters. Options: --json, --exclude 'tests/*', --strict (fail on warnings), --format github. In an agent session the same check is the check_telegram_project tool (local stdio server only, since it reads files).

On a GitHub repository, run it on every pull request; findings appear as annotations on the changed lines:

# .github/workflows/telegram.yml
on: [pull_request]
jobs:
  bot-api:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: Kay0k1/telegram-dev@v0.9.0
        with:
          path: .
          exclude: "tests/*"

Install

You need uv. No bot token, Telegram account or API key is needed: the server only reads public documentation.

MCP server and skill, one command

Run this in your project directory:

uvx telegram-dev install --agent claude    # or: codex, cursor

`--agent`	MCP server is added to	Skill is copied to
`claude`	`.mcp.json`	`.claude/skills/telegram-dev`
`codex`	`.codex/config.toml`	`.agents/skills/telegram-dev`
`cursor`	`.cursor/mcp.json`	`.cursor/skills/telegram-dev`

The installer keeps your other settings, backs up the files it changes and never overwrites an existing Telegram Dev setup. The server is pinned to the installed version, so nothing has to stay on disk besides the uv cache. Open a new agent session afterwards and approve the project MCP server when asked. Codex loads project settings only after you trust the project directory.

To upgrade later: uvx telegram-dev@latest install --agent claude --update.

Claude Code plugin

/plugin marketplace add Kay0k1/telegram-dev
/plugin install telegram-dev@telegram-dev-tools

This adds the skill and the MCP server for all your projects. Use either the plugin or the installer, not both.

MCP server only

claude mcp add --scope user telegram-knowledge -- uvx telegram-dev serve

codex mcp add telegram-knowledge -- uvx telegram-dev serve

Cursor:

Any other MCP client:

{
  "mcpServers": {
    "telegram-knowledge": { "command": "uvx", "args": ["telegram-dev", "serve"] }
  }
}

The tools work without the skill, but the skill is what makes an agent use them consistently.

Current docs for agents

Telegram ships a new Bot API version every month or two. A model's knowledge of it stops at its training cutoff, so agents confidently write code against fields that changed or never existed. They also mix up the Bot API, MTProto and TDLib, which have very different capabilities.

TON moves just as fast. The current TON docs call the coin Gram, describe Telegram Money, the wallet built into Telegram, and treat Tolk as the main contract language with FunC as legacy. On top of that, Telegram has its own rules: blockchain features in Mini Apps must use TON and TON Connect, and digital goods are sold for Stars.

With Telegram Dev the agent reads the current page before it writes code. A real Codex CLI run in a fresh project after uvx telegram-dev install --agent codex (tool calls trimmed):

> Use the telegram-dev skill. In a python-telegram-bot project, I want the inline
  'Delete' button to be red and show a custom emoji. Which Bot API fields do this,
  what are the requirements, and what does the request JSON look like?

  get_telegram_corpus_status {}
  search_telegram_docs {"query": "style icon_custom_emoji_id", "area": "releases"}
  search_telegram_docs {"query": "InlineKeyboardButton icon_custom_emoji_id style", "area": "bot-api"}
  get_telegram_document {"url": "https://core.telegram.org/bots/api", "offset": 167594, "length": 10500}
  get_telegram_document {"url": "https://core.telegram.org/bots/api-changelog", "offset": 39089, "length": 6000}

Use these InlineKeyboardButton fields:
- style: "danger" makes the button red.
- icon_custom_emoji_id is a custom emoji ID shown before "Delete".
- callback_data is your action identifier, 1-64 bytes.

The custom emoji requires either a bot that bought additional usernames on Fragment,
or a Premium bot owner with the message sent by the bot to a private chat, group or
supergroup. PTB 22.7+ supports both fields. Older Telegram clients show the button
unstyled. Source: https://core.telegram.org/bots/api#inlinekeyboardbutton

{"chat_id": 123456789, "text": "Choose an action:",
 "reply_markup": {"inline_keyboard": [[{"text": "Delete", "callback_data": "delete:42",
   "style": "danger", "icon_custom_emoji_id": "YOUR_CUSTOM_EMOJI_ID"}]]}}

The skill also knows when not to reach for a blockchain. Another real Codex run:

> Use the telegram-dev skill. Our Telegram Mini App sells a premium in-game skin
  for 2 TON. I want a Buy button that connects the user's wallet and charges them.

  get_telegram_document {"url": "https://core.telegram.org/bots/payments-stars", "refresh": true}
  search_telegram_docs {"query": "createInvoiceLink answerPreCheckoutQuery successful_payment XTR", "area": "bot-api"}
  get_telegram_document {"url": "https://docs.ton.org/applications/ton-connect/overview", "refresh": true}
  get_telegram_document {"url": "https://core.telegram.org/bots/webapps", "offset": 40310}

Use Telegram Stars (XTR) through the Bot API. The skin is a digital good, so
Telegram requires Stars and prohibits cryptocurrency payment for this purchase.
For permitted wallet features, use the TON Connect SDK.
1. Validate initData on the backend and bind the order to the user.
2. createInvoiceLink with currency "XTR", open it with Telegram.WebApp.openInvoice.
3. Answer pre_checkout_query, grant the skin once on successful_payment,
   store telegram_payment_charge_id for refunds.

Tools

Tool	What it does
`search_telegram_docs`	Full-text search over indexed sections, optionally limited to one area: `bot-api`, `mini-apps`, `telegram-api`, `tdlib`, `ton`, `releases`, `ton-releases`, `aiogram`, `python-telegram-bot`, `grammy` and others. Core Telegram docs rank first unless the query names a library
`get_telegram_document`	Reads an official page in pages of up to 40,000 characters; fetches it if it is missing or older than 6 hours
`get_telegram_corpus_status`	What is indexed, how fresh it is and which sources failed
`list_telegram_document_revisions`	Revisions of a page captured by this installation
`list_telegram_sources`	The 61 official entry points, grouped by API area
`list_telegram_cached_pages`	Pages already stored in the local cache, by area, with freshness
`get_bot_api_definition`	One Bot API method or type as structured data: parameters or fields, types, required flags, allowed values, length limits, and the changelog entries that mention it
`check_bot_api_call`	Checks a request your code sends against the current Bot API schema: unknown methods and parameters (with suggestions), missing required ones, wrong types, values outside the documented list, retired parameters with the version that replaced them
`list_bot_api_changes`	Official Bot API changelog as structured releases, filtered by version (`since_version="9.0"`) or by a method, type or field name
`check_telegram_project`	Checks every Bot API call in a local Python bot project in one run (see Check a bot project); local stdio server only

Checking Bot API calls

The Bot API tools are built from the official API page and changelog in the cache, so they follow Telegram's current version (Bot API 10.4 at the time of writing) without a hand-written schema. For example, check_bot_api_call on

{"method": "sendMessage", "params": {"chat_id": 1, "text": "Delete?", "reply_to_message_id": 42,
  "reply_markup": {"inline_keyboard": [[{"text": "Delete", "callback_data": "del", "style": "red"}]]}}}

returns an error for reply_markup.inline_keyboard[0][0].style (must be one of danger, success, primary) and a warning that reply_to_message_id was replaced by reply_parameters in Bot API 7.0. It does not call Telegram, so permissions, chat state and file contents still need a real test.

The cache starts empty. Pages are fetched the first time the agent reads them and indexed right away. To pre-fetch everything reachable from the configured sources:

uvx telegram-dev sync --max-pages 0
uvx telegram-dev search "sendVideoNote"
uvx telegram-dev status

What the skill covers

Area	What the agent checks
Bot API	Messages, keyboards, updates, webhooks, files, permissions
Formatting and emoji	Entities, UTF-16 offsets, custom emoji and who may use them
Media and stickers	Voice messages, video notes, sticker sets, file limits
Mini Apps	JS bridge, launch parameters, themes, server-side `initData` validation
MTProto and TDLib	User vs bot authorization, schema and layers, TDLib versions
Business and payments	Business connections, Stars, subscriptions, gifts, paid media
Groups and channels	Admin rights, forums and topics, moderation
TON	Telegram's blockchain rules for Mini Apps, Stars vs crypto, TON Connect and `ton_proof`, payment processing, jettons and NFTs (TEPs), Tolk contracts, testnet first, key safety
Bot libraries	aiogram, python-telegram-bot and grammY: the library docs that match the installed version, compared with the Bot API page; says when the indexed docs describe another version
Releases	Bot API changelog, API layers, TDLib changelog, TON node and Tolk changelogs, official announcements

The skill lives in skills/telegram-dev, with one reference file per area.

How fresh the docs are

Reading a page older than 6 hours re-fetches it. Search re-fetches up to six stale matching pages.
While the server runs, a background cycle checks up to 120 pages every 6 hours: entry points, newly found links, oldest pages.
If Telegram's site is unreachable, the cached text is kept and marked stale with the error, so the agent knows it is reading an old copy.
A daily GitHub Action checks all 61 official sources and publishes hashes and errors on the docs-status branch.

Settings such as the cache path (TELEGRAM_KNOWLEDGE_DB) and refresh intervals are in docs/runtime.md.

How much it helps

Measured with Codex CLI on 11 questions and 10 code-writing tasks (evals):

With Codex's own web access, answers were equally correct with and without Telegram Dev: the agent found the same official pages itself.
Without web access, answers stayed current with Telegram Dev (45/47 rubric facts against 38/47, 26/26 code checks against 23/26); plain Codex did not know the newest Bot API features or TON's current names.

So the documentation server matters most where the agent cannot browse. The project check works on your code rather than on a question, so it is useful either way. Run on open-source bots from GitHub, it mostly finds disable_web_page_preview and reply_to_message_id, retired in Bot API 7.0, which Telegram and the libraries still accept for compatibility. Details, including the review for false positives: docs/validation.md.

Self-hosting

The server also runs over Streamable HTTP, so a team can share one cache. A Dockerfile and a Caddy setup with HTTPS are included: docs/deployment.md.

uvx telegram-dev serve --transport streamable-http --port 8000

Limits

Search is local full-text search (SQLite FTS5), not semantic search. Exact method names and English terms work best. A small set of Russian aliases is built in.
Coverage is what can be reached from the 61 official entry points. TON pages come from the llms.txt index that docs.ton.org publishes for agents, so new TON pages are picked up automatically.
Bot libraries: guides and bot-facing references of aiogram (latest, built from its main branch), python-telegram-bot (stable docs and the official wiki) and grammY (English guides and plugins), about 250 pages. Generated per-type pages that mirror the Bot API, grammY's /ref/ and translations are not indexed; released aiogram and PTB versions (/en/vX.Y.Z/) can be read on demand. Telegraf, Telethon and Pyrogram are not indexed. Page history starts when your installation first fetched the page.
The server does not log in to Telegram, send messages, touch chats or hold wallet keys. It only reads public pages.
The skill improves how agents work. It does not guarantee a correct answer: check the code against Telegram before shipping.

Development

git clone https://github.com/Kay0k1/telegram-dev && cd telegram-dev
uv sync --frozen
uv run --frozen python -m unittest discover -s tests -v
uv run --frozen telegram-dev install --agent claude --project /path/to/test-project

CI runs the tests on Linux, macOS and Windows and checks the Docker image. docs/validation.md lists what has been verified by hand and what has not. Issues and pull requests are welcome; tests/behavioral-cases.json has agent scenarios that still need runs in each client.

License

MIT for the code and the skill. Fetched Telegram and TON documentation remains with its owners; the package does not bundle it (NOTICE). This is an independent project, not affiliated with Telegram or the TON Foundation.

telegram-dev

Documentation