Skip to main content

​
What is the Thor MCP server?

Thor’s Model Context Protocol (MCP) server exposes Thor’s capabilities to any MCP-compatible AI client. That includes Cursor, Claude Desktop, and custom integrations. Instead of building and maintaining your own multi-MCP data pipeline that stitches together GitHub, Linear, calendar, meeting transcripts, Slack and more, you can point your AI client at a single endpoint. Thor handles the cross-tool reasoning, permissions and context for you. The answers your coding agent or chat client gets are faster, cheaper, more accurate and more reliable.

​
What you can do with it

Once connected, your MCP client gets access to Thor’s core capabilities:
  • Ask Thor any question it can already answer in Slack. Issues, projects, documents, teams, meetings and decisions.
  • Pull the context pack for a GitHub or Linear issue. Parent work, related PRs, repo docs, meetings, and the decisions that actually bear on it.
  • Pull your daily standup summary (or a teammate’s, with the right permissions) directly into your editor or chat.
  • Add notes, blockers and updates to your standup without leaving your editor.

​
Connect your client

Cursor, Claude Code, and any other MCP-compatible client connect to Thor the same way: add a Thor entry to your client’s mcp.json configuration file pointing at your organization’s Thor MCP URL.
Replace <tenant> with app or your single-tenant subdomain name. If you’re not sure what your URL is, contact support@thor.ai.
Some clients, like Claude Desktop, don’t fully support standardized http/oauth connections yet. If you use Claude Desktop, or the config above doesn’t work for you, you can install mcp-remote and use the following config instead:
The first time your client connects, you’ll be redirected through an OAuth flow. Sign in with the Slack account you use for Thor. There are no API keys to manage.

​
Authentication and permissions

The MCP server resolves your identity from your OAuth login and applies the same organization and channel permissions as the Thor Slack app. You’ll only see data you already have access to in Thor.

​
Use cases

​
Your coding agents stop guessing

Someone decides something in a meeting. Someone else changes their mind in Slack on Friday. The ticket still says what it said three weeks ago. Nobody lied. The team just moved faster than its own notes. Then a coding agent reads that ticket and builds exactly what it says. Including the wrong parts. Thor keeps the record current and hands it to whoever picks up the work next, human or agent. When Thor has something worth adding, it injects a briefing into the GitHub or Linear issue body. That is the context pack, written onto the ticket so whoever opens the issue already has it. The block is folded so it does not clutter the description. A person sees one line, Additional context (powered by Thor AI). An agent reading the issue sees the full context pack: parent work, related PRs, governing docs, meetings, and decisions. The last line of that block tells the agent how to call Thor if the bullets are not enough.
Thor embeds that block into GitHub issue and Linear issue bodies. If you don’t see Additional context (powered by Thor AI) on your tickets yet, email support@thor.ai.
Thor doesn’t write the code. That’s what Cursor, Claude Code, and the rest are for. Thor makes sure whatever agents does the work has the right historical context.

​
Standup summaries and daily briefings

Skip the context switch to Slack. Ask your MCP client to fetch your standup or daily briefing directly:

​
Keep your project management system up-to-date

If you’re heads-down in your editor and need to update Linear with your progress, your MCP client can post it back to Thor: