# Overview of AI in TON (https://docs-fpm2731fy-ton-core-docs.vercel.app/llms/ecosystem/ai/overview/content.md)



There are AI products utilizing TON Blockchain, such as agents and wallet tooling. Additionally, TON documentation is AI-friendly, and enables all kinds of AI workflows.

## AI ecosystem [#ai-ecosystem]

### `@ton/mcp` [#tonmcp]

The main official entry point is [`@ton/mcp`](/llms/ecosystem/ai/mcp/content.md), a TON MCP server for agents. It exposes tools for balance checks, asset queries, transfers, TON DNS resolution, swaps, and [agentic wallet](/llms/ecosystem/ai/wallets/content.md) management. Use [`@ton/mcp`](/llms/ecosystem/ai/mcp/content.md) when an agent needs to operate on TON through MCP or through a skills-based setup.

For the catalog of official TON MCP servers, setup guides, and related skills, see the [TON MCP portal](https://mcp.ton.org/).

### Agentic wallets [#agentic-wallets]

[Agentic wallet contracts](/llms/ecosystem/ai/wallets/content.md) provide self-custody wallets for autonomous AI agents operating on TON. They are used by `@ton/mcp` in its default agentic wallets mode.

### AI skills [#ai-skills]

[Acton development skills](https://ton-blockchain.github.io/acton/docs/agent-skills/overview) are reusable instructions for coding agents that work with TON smart contracts. They tell the agent how to read TON documentation, which Acton commands to run, how to structure project files, how to generate [Tolk](/llms/tolk/overview/content.md) code, and how to migrate FunC projects to Tolk. All skills follow the [Agent Skills specification](https://agentskills.io/home).

### Community [#community]

The [AI Dev Wall on Telegram](https://t.me/ai_dev_wall) is a builder channel for TON AI tooling and project updates. This channel primarily features third-party, community resources.

## Documentation [#documentation]

TON documentation exposes several AI-facing surfaces:

* The contextual menu can copy page content, open a page in external AI tools, or connect to the documentation MCP.
* The [`llms.txt`](https://docs.ton.org/llms.txt) file provides a machine-readable index of documentation pages.
* Raw Markdown is available through `.md` page routes. This is useful when a tool needs the page content without the site UI. For example, to obtain raw contents of this page, query `https://docs.ton.org/ecosystem/ai/overview.md`.

{/* TODO(restore when available):
  - The in-site assistant can answer questions over the docs and cite relevant pages. Assistant can be invoked by a shortcut <kbd>Ctrl+K</kbd> (or <kbd>Cmd+K</kbd> on macOS), a query in the floating input field at the bottom of each page, from any selection on the page, or via a button on each code block to explain its contents.
  - MCP server allows reading and querying documentation pages via MCP clients: `https://docs.ton.org/mcp`. Unlike [`@ton/mcp`](/llms/ecosystem/ai/mcp/content.md), it does not provide agents with tools to act on TON, only the documentation content.
  */}