Rollouts data sources: every MCP fails with “can only run in local environments” (incl. official Sentry plugin)

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

  • Editor, Tab & Chat (autocomplete, Composer, in-editor agent)
  • Terminal & commands
  • Models, pricing & API keys (availability, Auto/Max, BYOK/Bedrock)
  • MCP & tools
  • Cloud Agents & Automations (scheduled/event)
  • BugBot & Code Review
  • Cursor CLI
  • Cursor Mobile
  • Remote (SSH / Dev Containers / WSL)
  • Account, billing & login
  • Something else…

Describe the Bug

Team: AUTOSIGNAL TECHNOLOGIES PTE. LTD. — Rollouts enabled, monitoring autosignal8/planner-swarm2.

Every data source we add under Automations → Rollouts → Settings → Data sources immediately shows “Failed to connect” with: “This MCP server can only run in local environments; Cursor’s backend does not connect to it or hold its sign-in.”

Reproduced with:

  1. Official Sentry plugin (https://mcp.sentry.dev/mcp, plugin-managed).
  2. Team mcp.json HTTP server o2-proxy (https://o2-mcp-proxy.auto-signal-net.workers.dev/mcp, static Authorization header). The server works from curl and returns tools/list; its request logs show no request from Cursor when the data source was created.
  3. Team mcp.json HTTP server openobserve (direct OpenObserve endpoint).

Steps to Reproduce

  1. Enable Rollouts for team AUTOSIGNAL TECHNOLOGIES PTE. LTD., monitoring autosignal8/planner-swarm2.
  2. Go to Automations → Rollouts → Settings → Data sources and add a data source, using any of the three MCP sources listed above (official Sentry plugin, o2-proxy HTTP MCP, or openobserve HTTP MCP).
  3. The new data source immediately shows “Failed to connect” with: “This MCP server can only run in local environments; Cursor’s backend does not connect to it or hold its sign-in.”
  4. Click “Try again”. The error is not re-evaluated.

Expected Behavior

Cursor’s backend should connect to a remote HTTP MCP (including an official plugin, and a Team mcp.json HTTP server with a static Authorization header) when it is added as a Rollouts data source, and should not treat it as a local-only MCP.

Screenshots / Screen Recordings

Screenshots attached: Rollouts Settings showing openobserve “Failed to connect”, and the Edit MCP server dialog with the same local-environments error. HTTP header values are empty in the screenshot and are not included here.

Operating System

This failure is in the Cursor Dashboard / backend (Automations → Rollouts), not in the desktop IDE, so the client OS does not appear to matter.

  • Windows 10/11
  • MacOS
  • Linux

Version Information

N/A for Cursor IDE / CLI — this happens when adding a Rollouts data source in the Cursor Dashboard, not in the desktop IDE or CLI.

For AI issues: which model did you use?

N/A — this is a Rollouts data-source connection failure, not an AI model issue.

For AI issues: add Request ID with privacy disabled

N/A — there is no chat Request ID; the failure occurs at data-source creation in the Dashboard.

Additional Information

list-change-monitor-mcp-connectors shows the error is written at creation (createTime == updateTime, e.g. changeMonitorMcpConnectors/13409109, created 2026-11-10T00:37:51Z), and “Try again” does not re-evaluate it.

Questions: Is a team setting (Privacy Mode, Cloud Agents MCP access, etc.) blocking backend MCP connections for Rollouts? How can we connect a remote HTTP MCP with a static header as a Rollouts data source?

Does this stop you from using Cursor?

2

Hi there!

We detected that this may be a bug report, so we’ve moved your post to the Bug Reports category.

To help us investigate and fix this faster, could you edit your original post to include the details from the template below?

Bug Report Template - Click to expand

Where does the bug appear (feature/product)?

  • Editor, Tab & Chat (autocomplete, Composer, in-editor agent)
  • Terminal & commands
  • Models, pricing & API keys (availability, Auto/Max, BYOK/Bedrock)
  • MCP & tools
  • Cloud Agents & Automations (cursor.com/agents, scheduled/event)
  • BugBot & Code Review
  • Cursor CLI
  • Cursor Mobile
  • Remote (SSH / Dev Containers / WSL)
  • Account, billing & login
  • Something else…

Describe the Bug
A clear and concise description of what the bug is.

Steps to Reproduce
How can you reproduce this bug? We have a much better chance at fixing issues if we can reproduce them!

  1. …
  2. …
  3. …

Expected Behavior
What is meant to happen here that isn’t working correctly?

Screenshots / Screen Recordings
If applicable, attach images or videos (.jpg, .png, .gif, .mp4, .mov)

Operating System

  • Windows 10/11
  • MacOS
  • Linux

Version Information

  • For Cursor IDE: Menu → About Cursor → Copy
  • For Cursor CLI: Run agent about in your terminal
IDE:
Version: 2.xx.x
VSCode Version: 1.105.1
Commit: ......

CLI:
CLI Version 2026.01.17-d239e66

For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)

For AI issues: add Request ID with privacy disabled
Request ID: f9a7046a-279b-47e5-ab48-6e8dc12daba1
For Background Agent issues, also post the ID: bc-…

Additional Information
Add any other context about the problem here.

Does this stop you from using Cursor?

  • Yes - Cursor is unusable
  • Sometimes - I can sometimes use Cursor
  • No - Cursor works, but with this issue

The more details you provide, the easier it is for us to reproduce and fix the issue. Thanks!

9

Hey @Robin_Some, this isn’t Privacy Mode or Cloud Agents MCP access - those servers are currently set to run only on members’ machines, so Cursor’s backend won’t connect to them at all (that’s why your proxy sees no request and “Try again” doesn’t change). It’s a team-level MCP setting that isn’t currently surfaced in your dashboard, so we’ll switch it off on your team’s side and I’ll follow up here once it’s done.

Your setup is otherwise correct: a URL + static Authorization header is the supported way to add a Rollouts data source, and it’ll connect once that’s cleared. One tip - Rollouts only runs read-only tools, so make sure the ones you want are exposed with readOnlyHint: true.

Quick check so we don’t break anything: are all your team MCP servers meant to be reached from Cursor’s backend (none only reachable inside your own network)? Your three look publicly reachable, so this should be straightforward.

10

Thanks @mohitjain — that explanation matches what we saw.

Yes: all of our team MCP servers are meant to be reached from Cursor’s backend. None of them are only reachable inside our own network, including the three in the original post.

Please go ahead and switch that setting off for our team. We’ll watch for your follow-up, and we’ll make sure the tools we want for Rollouts are exposed with readOnlyHint: true.