
TL;DR: Quire's first-party MCP server connects to any MCP-capable AI client — Claude Code, Claude Cowork, ChatGPT, Perplexity, Gemini CLI, or another — via a one-time OAuth handshake. Setup takes about five minutes, requires no API keys, and access is scoped to whatever your Quire user can already do.
Quire's MCP server lets you wire Claude Code, Cowork, or any MCP-capable AI client into your Quire workspace in about as long as it takes to reheat lunch. One OAuth handshake, and your AI assistant can create tasks, update statuses, and manage projects across your Quire workspace — no developer background required, no third-party glue, no Zapier workflows to maintain. If you have a Quire account and an MCP-capable AI client, you're already most of the way there.
Quire MCP works the same way in every MCP-capable AI client — Claude Code, Claude Cowork, ChatGPT, Perplexity, Gemini CLI, Cursor, Windsurf, or anything new since this post was written. The OAuth handshake, the supported actions, and the permission scope are identical; only each client's UI differs.
Before we get into steps — a quick preview of the destination. Once connected, you can hand Claude tasks like:
Every one of those goes from a 10-minute clicking exercise to a prompt. That's the whole reason to do this.
Three things to check before you start:
That's it. No API tokens to generate manually, no config files to edit unless you want to.
The pattern is the same everywhere — register the Quire MCP server in your client, authorize once with OAuth, verify with a prompt. Below, you'll find worked examples for Claude Code, Claude Cowork, ChatGPT, Perplexity, and Gemini CLI, plus a generic walkthrough that applies to Cursor, Windsurf, and every other MCP-capable AI client.
New to MCP or prefer a non-developer walkthrough? The Quire MCP guide covers the basics in plain language — what MCP is, what it does for you, and what to expect once it's connected.
For the full feature list, supported clients, and the current endpoint URL, see the Quire MCP listing in the Quire developers documentation.
Claude Code is the most direct way to wire Quire into a developer workflow — you can have Claude create tasks, update statuses, and pull project state from inside the terminal session you're already coding in.
/plugin to open the connector list."List my recent Quire projects" — and confirm Claude can read your workspace.That's the whole flow. The connection persists across Claude Code sessions until you revoke it.
For the latest setup details, supported actions, and version compatibility, see the Quire app for Claude in the Quire app directory.
Claude Cowork is Anthropic's collaborative workspace for Claude — it hides the MCP plumbing entirely, so the connector list looks like any other integration menu and the OAuth flow takes two clicks.
"What tasks are due this week in my Quire workspace?".Claude Cowork shows the connector status in Settings → Connectors at any time, so you can confirm the connection is live without leaving the app.
ChatGPT connects to Quire through its custom-connector system — once added, you can have ChatGPT pull project state, create tasks, and update statuses inside any conversation, just like any other built-in tool.
"What's on my Quire to-do list this week?" — confirm it can read your workspace.ChatGPT keeps the connector active across sessions until you remove it from settings.
Perplexity supports remote MCP servers as Custom Connectors on paid plans (Pro, Max, Enterprise) — once connected, you can ask Perplexity research questions that pull live data from your Quire workspace alongside its web sources.
perplexity.ai and go to Settings → Connectors.Quire), set the URL to https://mcp.quire.app/mcp, and choose Streamable HTTP as the transport with OAuth as the authentication method.Once connected, Perplexity can query your Quire data inside any conversation.
Custom MCP connectors require a paid Perplexity plan (Pro, Max, or Enterprise) — Quire itself works on every Perplexity plan; the limitation is on Perplexity's side. For setup details and version compatibility, see the Quire app for Perplexity in the Quire app directory.
Gemini CLI is Google's command-line interface for the Gemini family of models — once connected, you can have Gemini create tasks, update statuses, and pull project state from any terminal session, similar to Claude Code.
~/.gemini/settings.json (user scope) or .gemini/settings.json (project scope).Add the Quire MCP server under mcpServers:
{
"mcpServers": {
"quire": {
"httpUrl": "https://mcp.quire.app/mcp",
"authProviderType": "dynamic_discovery"
}
}
}
Restart Gemini CLI, then trigger OAuth with /mcp auth quire. dynamic_discovery makes the CLI auto-detect the OAuth requirement, register a client, and open a browser to authorize. (OAuth requires that your local machine can receive a redirect on http://localhost:7777/oauth/callback.)
Verify. Prompt with something like "List my recent Quire projects" and confirm Gemini can read your workspace.
The connection persists across Gemini CLI sessions until you revoke it.
For the latest setup details, supported actions, and version compatibility, see the Quire app for Gemini CLI in the Quire app directory.
If you're on a different MCP-capable client — a desktop AI app, an editor extension, or anything new since this post was written — the pattern is the same:
If your client supports the MCP registry, Quire should appear there directly. If it doesn't, you may need to add the server manually — check Quire's developer documentation for the current endpoint URL.
Any client that supports Streamable HTTP MCP servers with OAuth can connect — just point it at https://mcp.quire.app/mcp. For clients that only speak the older stdio transport, use a bridge such as mcp-remote:
{
"mcpServers": {
"quire": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://mcp.quire.app/mcp"]
}
}
}
Drop that block into your client's MCP config file, restart the client, and the OAuth handshake will trigger on first use.
The most useful way to build intuition is to try a few real prompts. These are starting points, not exhaustive — each one exercises a different capability.
"Summarize the current state of my 'Website redesign' project. What's done, what's in progress, what's stuck?"
Tests: reading tasks, understanding task status, synthesizing natural-language summaries from structured data.
"What tasks are assigned to me and due this week in the 'Procurement Approval' project? Sort by priority."
Tests: filtering, sorting, user context.
"What did I complete yesterday in Quire? What am I working on today?"
Tests: historical activity, current assignments, blocker detection.
"Here are my sprint planning notes: [paste notes]. Turn these into tasks in the 'Q3 Sprint' project, assigned to Brent and Bailey."
Tests: task creation, assignee inference, project selection.
"Mark all tasks in the 'Website redesign' project tagged 'ready-for-review' as complete."
Tests: tag filtering, batch operations, task updates.
"Draft a one-paragraph weekly status for stakeholders based on the last 7 days of activity in my active projects."
Tests: cross-project reading, summarization, tone matching.
"Create a new project called 'Onboarding — Alex' with these tasks: [list], assigned to Alex, due in the next two weeks."
Tests: project creation, task creation, assignment, due dates.
"Which tasks have been in 'In Progress' for more than 5 days? Tag them 'needs-attention' and add a comment asking for a status update."
Tests: filtering, tag management, comment creation.
"Copy the 'Monthly marketing launch' project structure into a new project called 'Marketing launch — August'."
Tests: project duplication, structure preservation.
"Are there any projects with milestones due this month where more than 30% of tasks are incomplete?"
Tests: milestone reading, percentage calculations, risk signaling.
Want the templates behind some of these workflows? The Quire templates library has sprint planning, onboarding, and marketing launch templates you can adapt.
Check the MCP connector status in your client's settings. If the connector shows as disconnected, re-authenticate. If it shows as connected but Claude still claims no access, try restarting your AI client — MCP connections sometimes need a reload to register.
This is usually a naming ambiguity — if you have three projects with "Marketing" in the name, the AI will pick one. Either use more specific project names in your prompt ("Marketing Q3 — launch campaign") or explicitly disambiguate by quoting the full project name or pasting the project URL.
Some enterprise environments block OAuth redirects in the browser. If you get stuck mid-flow, check your firewall/proxy settings — the redirect URL needs to be reachable.
Quire's MCP respects your user permissions — if there are projects you don't have write access to, the AI won't either. For tighter control within a project, use a Quire role with restricted permissions for the user account you're connecting.
If you want to disconnect — temporarily or permanently:
/plugin.Revoking from Quire's side is the most authoritative — the AI client will lose access immediately, regardless of its local state.
Connecting Claude to Quire takes five minutes, requires no code, and uses scoped OAuth that you can revoke at any time. Most of the setup time is authentication — the protocol itself does the heavy lifting of translating between your AI prompts and Quire's API.
Once connected, the value shows up fastest in the coordination work you were already doing — status updates, sprint planning, task creation from meeting notes, and triage. Start with one or two of those workflows and build from there. Not everything needs to be AI-assisted, but the handful that benefit most will give you back real hours.
Quire's MCP server works with any MCP-capable AI client — Claude Code, Claude Cowork, ChatGPT, Perplexity, Gemini CLI, and more. Open your client's MCP settings, add the Quire MCP server, authenticate via OAuth, and verify with a test prompt. About five minutes total.
Yes. Add it through Claude Code's /plugin menu or your MCP config file, authenticate once, and Claude Code can read and write Quire data alongside your normal coding workflow.
Yes. In Cowork, go to Settings → Connectors → Add Connector, pick Quire, and sign in. Cowork can then manage tasks, projects, and documents without leaving the chat.
Any MCP-capable client — Perplexity, Gemini CLI, Cursor, Windsurf, Zed, Codex, and others. The OAuth-based setup pattern is the same in every client. See Quire's developer documentation for the current list of tested integrations.
Anything your Quire user account can do — create and update tasks, manage tags and sublists, read and write comments, edit documents, pull insights, and move work between projects.
Yes. Scoped OAuth, revocable anytime from Quire's account settings. The AI only sees what your user can see, and all actions are logged to your account.
Ready to run your project management from your AI assistant?
Start your free trial at quire.io/signup — no credit card, full access, 30 days.