Read Slack Messages with AI (Claude, Cursor, VS Code) — No Tokens, No Slack App

To give Claude, ChatGPT, Cursor or VS Code access to Slack on a Mac, install LMCP — a free, local MCP server for macOS. Instead of going through Slack's Web API, it reads directly from the local cache the Slack desktop app already keeps on your machine, so there is no Slack app to create, no OAuth scopes to pick, and no workspace-admin approval to wait on. Your AI can then list your workspaces, browse channels, read message history, and search across everything — enough to catch up on a busy channel or find a decision from three weeks ago in one prompt. Setup takes about two minutes. This is read-only: LMCP cannot post, edit, or delete anything in Slack.

Why Give Your AI Access to Slack?

Slack is where a lot of real decisions get made, but it is also the app people fall behind on fastest — a few days away, or one day of back-to-back meetings, and a busy workspace has piled up hundreds of messages across a dozen channels. Scrolling to find the handful of things that actually need you is slow, and it is exactly the kind of tedious summarization work an AI is good at.

With LMCP, your AI assistant can read your Slack channels and DMs directly and answer in plain English: “What did the team decide about the launch date?” or “Catch me up on #engineering since Friday.” It reads through the messages, pulls out what matters, and gives you a summary instead of a scrollbar.

Because the same server also reads email and Microsoft Teams, your AI can connect a Slack conversation to the email thread or Teams chat that continued it — something none of those apps can do on their own.

What You Will Need

• macOS 12 (Monterey) or later — Apple Silicon or Intel
• Any MCP-compatible AI client: Claude Desktop, Cursor, VS Code (Copilot), Windsurf, Zed, or ChatGPT
• The Slack desktop app, installed and signed in to at least one workspace

You do not need a Slack app, OAuth scopes, a bot token, or workspace-admin approval. LMCP reads the same local data the Slack desktop app already keeps on your Mac, so every workspace you are signed into there is automatically available to your AI.

Step 1: Install LMCP on Your Mac

Download LMCP and install it:

1. Open the downloaded LMCP-Installer.pkg from your Downloads folder
2. Follow the installer — it installs Local MCP to Applications for you (no dragging)
3. Open Local MCP from your Applications folder — it appears in your menu bar

Takes about 30 seconds. Your AI clients are configured automatically.

Make sure the Slack desktop app is installed and signed in before you ask your AI anything — LMCP reads Slack's local data, so Slack has to have synced it first. The first time a Slack tool runs, macOS will ask you to grant LMCP Full Disk Access; click “Allow” — that is what lets LMCP read the local cache at all.

Step 2: Restart Your AI Client

Quit your AI client completely (Cmd+Q, not just closing the window) and reopen it. It needs to restart to discover the new MCP tools.

• Claude Desktop — look for the hammer icon in the chat input
• Cursor — MCP tools appear automatically in the composer
• VS Code — check the MCP panel in Copilot settings, or reload the window (Cmd+Shift+P → “Reload Window”)
• ChatGPT — add https://local-mcp.com/mcp as a custom connector

Step 3: Try It Out

Start a new conversation and try these prompts:

• “List my Slack workspaces” — confirms LMCP can see your Slack data before you ask anything harder.
• “Catch me up on my Slack channels from today” — summarizes recent activity across everything you have open.
• “Search my Slack for messages about the Q3 budget” — finds the relevant thread by keyword instead of you scrolling for it.
• “What did the team say in #engineering this morning?” — reads and summarizes one specific channel.
• “Did anyone mention me with something I need to act on?” — scans recent messages for action items directed at you.

Every one of these is a read: your AI looks at what is already in Slack and reports back. There is nothing to confirm or approve, because nothing changes.

Which Slack Workspaces Work?

Any workspace you are signed into in the Slack desktop app is available — LMCP does not connect to a specific workspace, it reads whatever the app itself has already synced:

• Multiple workspaces — if you are signed into several, your AI can read and search across all of them, or you can point it at one by name.
• Public and private channels — anything visible to your account in the Slack app is visible to LMCP; it inherits your own permissions, nothing more.
• Direct messages and group DMs — these live in the same local cache and are readable the same way as channels.

The trade-off for skipping Slack's API is that LMCP depends on what the desktop app has actually cached, not a live query against Slack's servers:

• Only synced history is available — a channel you have not opened in months may only be partially on disk.
• The desktop app has to be installed — the browser version of Slack keeps no local database to read.
• Freshness follows sync — keep Slack open so the cache stays current.

It is the same data the Slack app shows you when you open it, just read by your AI instead of your eyes.

What Your AI Can Do with Slack

LMCP exposes four Slack tools, and all four are read-only:

• List workspaces — every Slack workspace you are signed into on this Mac
• List channels — browse channels and DMs within a workspace
• Read channel messages — pull recent message history from a specific channel
• Search messages — find messages across your channels by keyword or topic

That is the full list. There is no tool to post, edit, delete, or react to a Slack message — if you ask your AI to send something to Slack, it cannot, and it should tell you that plainly rather than pretend. If your workflow genuinely needs your AI to post on your behalf, see the comparison further down for the token-based alternatives that can.

Four read tools still add up to a capable research layer: catching up on channels, finding a decision, and pulling context into a task somewhere else — Mail, Reminders, OmniFocus, wherever the follow-up belongs.

Privacy: Everything Stays on Your Mac

This is the same architecture as every other app LMCP connects to:

• LMCP runs as a local server on localhost:8765
• It reads the Slack desktop app's local database directly, opened read-only
• Message content goes directly from that local database to your AI client — no intermediate servers
• No Slack data is stored, cached, or logged by LMCP itself

No Slack app, no OAuth flow, no bot token sitting in a config file that could leak. Your messages stay exactly where they are — in Slack's own local data on your machine. GDPR and CCPA compliant by architecture.

Technical Deep-Dive: How LevelDB Parsing Works

For the technically curious, here is how LMCP actually gets Slack data off disk.

The Slack desktop app stores its IndexedDB data on macOS at paths such as:

~/Library/Application Support/Slack/IndexedDB/*.leveldb
~/Library/Containers/com.tinyspeck.slackmacgap/Data/Library/Application Support/Slack/IndexedDB/*.leveldb

This is a Chromium-style LevelDB database — the same format Chrome and other Electron apps use to back IndexedDB. LMCP parses it with the open-source ccl_chromium_reader library (originally built for digital forensics), which understands Chromium's V8 serialization, LevelDB's append-only log and compaction, and IndexedDB's object stores. The parsed records are filtered into clean JSON — channels, messages, senders, timestamps — that your AI assistant can work with.

The database is opened read-only. LMCP cannot corrupt or modify your Slack cache, and none of the four Slack tools has a code path that writes back to it.

LMCP vs Other Slack MCP Servers

Searching for a “Slack MCP server” also turns up two other serious options: the community project korotovsky/slack-mcp-server (a token-based server with a “stealth mode” that reuses browser session tokens) and Slack's own official hosted MCP server at mcp.slack.com. Both are good at what they do — the honest comparison is what each requires:

In short: if you need your AI to post to Slack, korotovsky's server or the official one are the right tools — both with token or admin prerequisites. If you want read and summarize with zero setup, next to Teams, Mail and iMessage in the same server — that's LMCP.

Troubleshooting

AI says it cannot find any Slack data

Make sure the Slack desktop app — not the browser version — is installed and you are signed in. Open Slack, let it load a channel or two so it writes to its local cache, then ask your AI again.

A workspace or channel is missing

LMCP only sees what the desktop app has already synced. If a workspace was added recently, or a channel has not been opened in a long time, switch to it in Slack for a moment so the app pulls it down, then retry.

Search returns nothing, or feels out of date

Search runs against the local cache, not a live query against Slack's servers. If Slack has been closed for a while, open it and let it sync before searching again; very old messages never opened locally may not be indexed at all.

MCP tools not visible in AI client

Make sure you fully quit and restarted your AI client. You can also check the config file:

• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
• Cursor: ~/.cursor/mcp.json
• Windsurf: ~/.codeium/windsurf/mcp_config.json

Permission denied errors

Reading another app's local data on macOS requires Full Disk Access. Go to System Settings > Privacy & Security > Full Disk Access and enable Local MCP. If it is not listed, download and reinstall LMCP.

Real-World Workflows

The value isn't reading one message — it's having your AI synthesize across channels, and across the other apps LMCP also reads.

Catching up after time away

Instead of opening twenty channels and scrolling, your AI reads them all and hands you one prioritized digest — the catch-up a vacation used to cost you an afternoon.

Finding a decision buried in a channel

This is the search tool doing what search is for: a decision from weeks ago, found and summarized in one prompt instead of a scroll-and-Ctrl-F hunt.

A daily digest of mentions across Slack, Mail and Teams

Because LMCP reads Slack, Mail and Teams in the same server, your AI can build the single “what needs me today” list that none of those apps produces on its own. See catching up across Teams, Slack and WhatsApp for a longer version of this workflow.

Turn Slack threads into tasks

Commitments hide in chat — “I'll send that over”, “let me check”. Your AI surfaces them and, with LMCP's Reminders and OmniFocus tools, turns them into tasks you can actually track.

What Else Can LMCP Do?

Slack is one of roughly two dozen app domains LMCP connects your AI to, alongside email, Microsoft Teams, Calendar, Contacts, OneDrive, and local files. Check the full list of guides or visit local-mcp.com for all 188+ tools.

Frequently Asked Questions

Can I connect an AI to Slack without a bot token or admin approval?

Yes — if you only need read access. LMCP reads the local cache the Slack desktop app already keeps on your Mac, so there is no app to create, no token to extract, and nothing for an admin to approve. Every API-based option (including “stealth” token servers) still depends on tokens tied to your workspace.

Can my AI search Slack messages?

Yes. Search is one of the four Slack tools LMCP exposes, and it runs across every channel and DM in your local cache by keyword or topic — useful for finding a specific message or decision without knowing which channel it was posted in.

Can my AI send messages to Slack?

No. LMCP's Slack tools are read-only: list workspaces, list channels, read channel messages, and search messages. There is no tool to post, reply, edit, or react in Slack. If posting is what you need, see the comparison above for token-based servers that support it.

Does this work with multiple workspaces and private DMs?

Yes to both. LMCP reads every workspace you are signed into — your AI can search across all of them or focus on one by name — and it can read whatever your own account can already see, including direct messages and group DMs, since those live in the same local cache as channels. It grants no access beyond your existing Slack permissions.

Which Slack MCP server should I choose?

Choose korotovsky/slack-mcp-server if you need posting and are comfortable managing session tokens. Choose Slack's official MCP server if your workspace has enabled it. Choose Local MCP if you want zero-config read access to Slack alongside Teams, Mail, iMessage and WhatsApp — including from ChatGPT and Claude.ai on the web.

Ready-Made Recipes

Copy-paste workflows that use the Slack tools on this page:

• Find every unanswered client request across WhatsApp, Slack, Teams & email
• Have your AI prep you for your next meeting

Related Guides

• Catch Up on Teams, Slack and WhatsApp with One AI Summary
• Read Microsoft Teams Messages with AI — No Graph API
• AI + Email on Mac: Full Integration Guide
• Best MCP Server for Mac Productivity in 2026
• Read Slack Messages with AI on Windows — No Tokens, No Slack App