A project management tool that works with Claude and ChatGPT is one that offers an MCP (Model Context Protocol) server, so the assistant can read your projects and tasks and create or update them from the chat. Most established tools now offer one, including Asana, ClickUp, Notion, monday.com and Linear. The real differences are what the assistant can do through it, who can set it up on your plan, and how much company context sits behind the tasks.
Why connect your project tool to an AI assistant
Without a connection, ChatGPT and Claude start every conversation knowing nothing about your work. You paste in a task list, ask for a summary, then copy the result back. With a connection, you can ask "What's overdue in the launch project?" or "Turn these meeting notes into tasks for Sara and Tom" and the assistant does it directly in your tool.
The standard behind this is MCP. Anthropic introduced it in November 2024 as an open standard for connecting AI assistants to the systems where data lives, replacing one-off integrations with a single protocol (Anthropic, 2024). See What is MCP? for a plain explanation.
How it works in Claude and in ChatGPT
Claude. You add a remote MCP server as a custom connector by entering its URL and signing in. Custom connectors are available on Free, Pro, Max, Team and Enterprise plans, with Free limited to one; on Team and Enterprise an owner adds the connector for the organization and members connect individually (Anthropic, 2026).
ChatGPT. Custom MCP server connections with full MCP support, including write actions, are available on ChatGPT Business and Enterprise/Edu on the web. Admins and owners add and test a server, then publish it for the workspace (OpenAI, 2026). Some project tools, such as Asana, also offer a ready-made app in ChatGPT, which avoids the custom setup (Asana Developers, 2026).
In both, the assistant can only do what your account in the project tool allows. Most project tool MCP servers sign you in with OAuth.
Which project tools offer an official MCP server
Checked on each vendor's own site in November 2026.
| Tool | Official MCP server | Clients named by the vendor | Notes |
|---|---|---|---|
| Asana | Yes | Claude, ChatGPT, Claude Code, Cursor, VS Code and other clients | Create, assign and report on work; older beta endpoint deprecated |
| ClickUp | Yes, all plans | "All major MCP clients" | Tasks, time tracking, status reports; no deletion tools by design; daily rate limits per plan |
| Notion | Yes | Claude, ChatGPT, Cursor and any MCP client | Search, create and edit pages, manage databases |
| monday.com | Yes, all plans | Claude, ChatGPT, Copilot, Cursor | Installed by an admin from the marketplace; can limit to specific workspaces |
| Linear | Yes | Claude, Claude Code, Cursor, VS Code, Codex and others | Issues, projects, comments; read-only endpoint available |
| Orakis | Yes | Claude, ChatGPT and other MCP assistants | Goals, tasks, documents and knowledge in one workspace |
Sources: Asana Developers, ClickUp Developer Docs, Notion Help Center, monday.com and Linear Docs (all 2026).
What to look for
1. Can the assistant write, or only read? Reading is useful for summaries. Writing (creating tasks, updating status, assigning) is where the time savings are. Check whether you also want a read-only option for some users; Linear, for example, offers a read-only endpoint (Linear Docs, 2026).
2. Does the assistant get the context behind the tasks? A task list alone answers "what is late?" but not "why did we change the deadline?" or "what do we usually charge for this?" The answer improves when the same connection also reaches documents, decisions and company knowledge. See how to give ChatGPT or Claude access to company knowledge.
3. Who controls access? On team plans, an admin usually adds and approves connectors. Look for scoping (only some workspaces or projects) and a clear record of what the assistant changed.
4. Safety defaults. Anthropic advises connecting only servers you trust, reviewing requested permissions and disabling write tools you don't need, because malicious servers can hide instructions in their output (Anthropic, 2026). Defaults like ClickUp's choice to leave out deletion tools help.
MCP connection vs. built-in AI
Connecting Claude or ChatGPT to your project tool is different from the AI that lives inside the tool:
- Through MCP, the assistant works when you ask it to. It does not watch the project while you are away.
- Built-in AI can run in the background: notice a late task, follow up with the owner or prepare a daily briefing.
Many teams want both: their favorite assistant for ad-hoc questions and drafting, and a project tool that keeps coordinating on its own. See AI tool that follows up on tasks.
How to set it up in 15 minutes
- Check your plans. Confirm your Claude or ChatGPT plan supports custom connectors, and whether an admin must add them.
- Find the vendor's MCP URL in its developer docs, or the ready-made app in your assistant's directory.
- Connect and sign in with OAuth. Limit scopes or workspaces where you can.
- Test with read questions first: "What's due this week in project X?"
- Then try one write action, such as creating a task, and check it in the project tool.
- Agree on rules with the team: which actions the assistant may take, and which need a person.
How Orakis does it
Orakis offers MCP so Claude, ChatGPT or any other MCP-capable assistant can work directly with your Orakis workspace. Because goals, tasks, team chat, documents and company knowledge live in the same workspace, the assistant gets the context behind the work, not just a task list. Inside Orakis, the built-in agent Alfred and the background coordinator keep following up while you are not in a chat. Leads decide how much Orakis may approve on its own, and every approval is visible. Orakis is a young product in closed beta, with fewer integrations than the established tools.
FAQ
Do I need a paid ChatGPT or Claude plan? For Claude, custom connectors are available on all plans, with Free limited to one. For ChatGPT, custom MCP servers require a Business or Enterprise/Edu workspace. Plans change, so check the current help pages.
Is it safe to let an AI assistant change my tasks? It is as safe as the permissions you give it. Start with read access, add write actions one at a time, and use tools that log what the assistant did.
Can I use Claude and ChatGPT on the same project? Yes. MCP is an open standard, so the same project tool server can be connected to several assistants at once.
Sources
- Introducing the Model Context Protocol, Anthropic (2024)
- Get started with custom connectors using remote MCP, Anthropic (2026)
- Connecting custom MCP servers in ChatGPT, OpenAI (2026)
- Using Asana's MCP server, Asana Developers (2026)
- Connect an AI assistant to ClickUp's MCP server, ClickUp Developer Docs (2026)
- Notion MCP, Notion Help Center (2026)
- monday MCP, monday.com (2026)
- Linear MCP server, Linear Docs (2026)