Source: https://docs.slack.dev/ai/build-with-ai

# AI-assisted development

Developers using an AI coding assistant, IDE, or a chat-based LLM to assist in their Slack app building journey have several ways to give their tools better context about the Slack platform.

* * *

## Install the Slack MCP server {#mcp-server}

The [Slack MCP server](/ai/slack-mcp-server) gives your AI tool direct access to your Slack workspace data. Once connected, your AI assistant can search channels, read messages, and use real conversation context to help you build.

Set up instructions are available for [Claude](/ai/slack-mcp-server/connect-to-claude) and [Cursor](/ai/slack-mcp-server/connect-to-cursor), among other MCP-compatible clients.

* * *

## Point your AI tool to llms.txt {#llms-txt}

Add a rule to your AI tool's project instructions (for example, a `CLAUDE.md` or `.cursorrules` file) telling it to start with our documentation index:

```
For Slack platform documentation, start with https://docs.slack.dev/llms.txt
```

This file provides a structured overview of the platform, designed for LLM consumption. A full index of every documentation page is available at [docs.slack.dev/llms-sitemap.md](https://docs.slack.dev/llms-sitemap.md).

* * *

## Ingest the docs in bulk {#llms-full}

The complete docs site is concatenated into an [`llms-full.txt`](https://docs.slack.dev/llms-full.txt) file. We also offer language-specific bulk files.

File

Contents

[`llms-full.txt`](https://docs.slack.dev/llms-full.txt)

Everything

[`llms-full-platform.txt`](https://docs.slack.dev/llms-full-platform.txt)

Slack platform docs, reference, and language-agnostic tools (Slack CLI, GitHub Action)

[`llms-full-js.txt`](https://docs.slack.dev/llms-full-js.txt)

Bolt for JavaScript and the Node Slack SDK

[`llms-full-python.txt`](https://docs.slack.dev/llms-full-python.txt)

Bolt for Python and the Python Slack SDK

[`llms-full-java.txt`](https://docs.slack.dev/llms-full-java.txt)

Bolt for Java and the Java Slack SDK

[`llms-full-deno.txt`](https://docs.slack.dev/llms-full-deno.txt)

Deno Slack SDK

Pair a language file with `llms-full-platform.txt` for platform concepts, authentication, and API reference.

These files are sized for indexing and retrieval pipelines, not for a single prompt. You can drop individual pages into prompts as described in the [access pages as markdown](#markdown) section.

* * *

## Access pages as markdown {#markdown}

Every documentation page on this site is available as raw markdown by appending `.md` to the URL. For example:

```
https://docs.slack.dev/quickstart → https://docs.slack.dev/quickstart.md
```

This is useful for:

*   Copying page content into a prompt for context
*   Agents that fetch URLs dynamically (MCP `fetch` tool, Claude Code's WebFetch, etc.)
*   Building up a local reference for your project

You can also use the **Copy as Markdown** button at the top of any page to copy its contents to your clipboard.

* * *

## Use the "Open in..." feature {#open-in}

Each documentation page has an "Open in..." feature within the "Copy as markdown" button drop-down that lets you send the page directly to ChatGPT, Claude, or Perplexity as context for a conversation.

* * *

## Use the Slack CLI {#cli}

AI tools and agents can run [Slack CLI](/tools/slack-cli) commands directly in a terminal. The Slack CLI supports app scaffolding and local development.

The `slack docs` command searches Slack API documentation from the command line, which is useful for agents that can execute shell commands.

* * *

## Tips for better results {#tips}

When prompting an AI tool about Slack development:

*   **Specify your framework or SDK.** Mention whether you're using the [Bolt framework](/tools/#bolt) (and which language) or one of the [SDKs](/tools/#sdks), since implementation details differ significantly.
*   **Reference specific API methods.** Instead of "send a message," say "use `chat.postMessage`" to get more accurate guidance.
*   **Provide your app's manifest or scopes.** Sharing your current app configuration helps the LLM suggest relevant next steps without guessing.
*   **Link to the relevant docs page.** Paste the `.md` URL of the page you're working from so the AI tool can fetch it for full context.