Integrations

Connect AI & the MCP Server

Where the Connect AI page lives, who can open it, what it gives you, and where the full MCP Server documentation is.

DevStride ships a built-in Model Context Protocol (MCP) server at https://api.devstride.com/mcp. Connect Claude, ChatGPT, Cursor, Copilot, Codex, or Gemini CLI to it and your assistant can read and change your DevStride work in plain language, with exactly the permissions you already have.

Connecting is a personal action: whatever you connect acts as you, within the permissions you already have. The Connect AI page lives under Configure Settings → AI Assistants → Connect AI and is shown to people whose role grants Manage Ai Assistant. It opens Connect your AI client (MCP), which shows the MCP Server URL, lets you choose OAuth 2.1 (recommended) or API Key, and gives a ready-to-paste setup snippet for each client.

Members without that permission connect the same way, without the page: Setting Up Your Connection has the server URL, the snippet for every client and the steps, and the same panel, pre-filled for the API-key method, opens for every member from My Account → API Keys → New Key. The AI Assistant's greeting also links to that guide (connect an external agent); for an admin it opens the Connect AI page too.

What administrators should know

  • A connection acts as the person who made it. There are no read-only scopes yet, so an assistant connected by an Owner can do anything an Owner can. Encourage people to connect from accounts with the access they actually need.
  • OAuth connections appear on each member's My Account → Connected Apps page, where they can revoke them. A connection lasts at most 90 days and stops renewing when the member leaves, is signed out, or the organization no longer allows the way they signed in when they approved it; see How long a connection lasts. API keys appear under My Account → API Keys; an administrator whose role grants MANAGE_ALL_API_KEYS sees every member's keys there — see API Keys.
  • All API keys and MCP connections in an organization share one rate limit of 1,000 requests a minute.