[BUG] Remote triggers fail to load OAuth MCP connectors in execution environment

Status Fixed / completed
Reported on v2.1.89
Maintainer reply ✓ Yes — ashwin-ant
Activity 7 comments · opened Apr 1, 2026 · closed Apr 18, 2026
💡 Likely answer: A maintainer (ashwin-ant, collaborator) responded on this thread — see the highlighted reply below.

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Remote triggers with OAuth MCP connectors (Slack, Atlassian) stopped working on 1 April 2026.
Every scheduled and manual run fails — the MCP tools are reported as unavailable in the execution environment. The same connectors work correctly in a local Claude Code session, confirming the tokens are valid and the issue is specific to the remote trigger environment.

What Should Happen?

The remote trigger execution environment should load the OAuth credentials for configured MCP connectors and make those tools available, the same way they work in a local session.

Error Messages/Logs

"The required Slack and Jira MCP tools are not available in this environment."

Steps to Reproduce

Steps to reproduce

  1. Configure a remote trigger with one or more OAuth MCP connectors (e.g. Slack via mcp.slack.com/mcp)
  2. Add the connector UUIDs to the trigger's mcp_connections
  3. Run the trigger (manually or via cron)
  4. The MCP tools are not available in the execution environment

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.1.89 (working on 31 March 2026, broken on 1 April 2026 — no version change between runs)

Claude Code Version

2.1.89

Platform

Other

Operating System

macOS

Terminal/Shell

Other

Additional Information

Platform: Claude Code remote triggers (claude.ai)
Terminal/shell: zsh

Connector details:

Troubleshooting already attempted:

  • Updated all connector tool permissions to "always allow" — no effect
  • Disconnected and reconnected both connectors in claude.ai — no effect
  • Created a brand new remote trigger with the same connector UUIDs — same error

The fact that these connectors work in a local session but not in the remote trigger
environment points to the remote execution environment failing to inject OAuth credentials
at runtime.

View original on GitHub ↗

7 Comments

rorynickolls-skyral · 5 months ago

I've been seeing this issue this morning too.

Rocks-D-xebec-0 · 5 months ago

same issue from my side

brettjenkins · 5 months ago

I've had this issue today too

superk9user · 5 months ago

same issue and other connectors as well.

aidanmoonama · 4 months ago

Workaround: Use Agent subagents to call MCP tools

The main session can't access MCP connector tools, but spawned Agent subagents can call them directly by name. See #43320 for full details and prompt template.

Main Session (no MCP access)
  ├── spawn Agent → mcp__Slack__slack_send_message         ✅
  ├── spawn Agent → mcp__Notion__API-query-data-source     ✅
  ├── spawn Agent → mcp__Google-Calendar__gcal_list_events  ✅
  ├── (collect all results)
  └── spawn Agent → mcp__Slack__slack_send_message          ✅

Add this to your trigger prompt:

MCP tools will NEVER appear in ToolSearch. Do NOT use ToolSearch,
sleep/retry, or direct MCP calls from the main session.
The ONLY way to call MCP tools is via Agent subagents.

Burns extra tokens but works consistently. Tested with Slack, Notion, Google Calendar, Gmail, and Granola connectors.

ashwin-ant collaborator · 4 months ago

This was fixed in v2.1.101 — Remote-trigger and headless sessions now pick up OAuth MCP connectors that finish connecting after the turn has started. If you're still seeing this in the latest version, please comment with your version and repro and we'll reopen.

github-actions[bot] · 4 months ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.