There's another community post with 404 debugging tips that might be helpful. Please give these solutions a try and let us know how it goes.

https://community.vercel.com/t/debugging-404-errors/437

A human should be around soon to offer more advice. But you can also get helpful information quickly by asking v0.

I can confirm the same issue with the official Vercel MCP integration in ChatGPT.

OAuth completes successfully and list_teams correctly returns my authorized team.

However:

• list_projects returns an empty array
• get_project using the exact Project ID returns 404 Not Found
• the project exists and is accessible normally in the Vercel Dashboard
• the project is explicitly selected in the OAuth authorization configuration

I also reconnected the integration, reauthorized the team/project, completed “Allow Access”, refreshed ChatGPT, and verified the exact Team ID and Project ID from the Vercel Dashboard.

The result remains unchanged.

This appears to be an OAuth project-scope / MCP mapping issue rather than a deployment 404.

No deployments or write operations were performed during testing.

Thanks for the automated guidance. Could a human from the Vercel MCP team investigate this project-access discrepancy?

The official Vercel integration in ChatGPT identifies my Pro team, but list_projects returns [], get_project returns 404, and list_deployments returns 403. The signed-in dashboard shows the project. Reauthorization has not resolved this.

Only one deny-only bootstrap submission was made, containing no consultation application, credentials or customer data. Its initial receipt reported INITIALIZING, target=preview; the support assistant later reported READY on the production target. That later statement remains unverified. Passing Attempt 1 source checks do not establish the deployment’s actual state.

Could your team please:

1. Verify the MCP OAuth principal’s team/project access mapping and explain the inconsistent reads.
2. Reconcile that same original deployment against authoritative records: ownership, requested versus actual target, any target-change history, build/source identity, aliases and protection.
3. Provide an approved private channel for identifiers, the support conversation and redacted diagnostic evidence, plus a read-only verification step.

The AI > Vercel MCP support form directed us here because this Beta product receives community-only support. Please route this to a human MCP owner and provide a tracking reference.

This request authorizes read-only investigation only. No second deployment or recreation, promotion, domain/alias changes, protection changes, or Production changes are authorized. Production remains No-Go.

Bug Description

The Vercel MCP integration inside ChatGPT is crashing immediately after a successful OAuth handshake when attempting to fetch project metadata for a Pro Team workspace.

The connector successfully authenticates and identifies the Pro workspace. However, before it can return the project list (via list_projects), the integration becomes completely unavailable/disconnected inside ChatGPT. This appears to be a token/project-scope mapping issue on the remote protocol server (://vercel.com).

Steps to Reproduce

1. Connect the Vercel MCP integration in ChatGPT.
2. Complete the OAuth handshake successfully.
3. Attempt to run a command that triggers a project or deployment lookup (e.g., listing projects or checking recent builds).
4. Provide the exact Team ID when prompted.
5. The integration hangs, fails to return the project list, and terminates the connection entirely.

Environment & Metadata

• Integration: Official Vercel MCP Server (://vercel.com)
• Client: ChatGPT Plus / Team
• Account Type: Vercel Pro
• Affected Team ID: team_yBO3XevZeo9wqBO9d5drZGOX
• Observed Behavior: Connector crashes andteam_yBO3XevZeo9wqBO9d5drZGOXteam_yBO3XevZeo9wqBO9d5drZGOX becomes "unavailable" right before listing projects.