[BUG] CCR scheduled triggers fail to load MCP connector tools in unattended runs

Status Closed — not planned
Reported on v2.1.89
Maintainer reply ✓ Yes — ashwin-ant
Activity 6 comments · opened Apr 1, 2026 · closed Apr 5, 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?

Claude.ai connectors attached to a scheduled remote trigger load MCP tools
successfully during manual/test runs but fail with "No MCP tools are loaded
in this session" when the same trigger fires on a cron schedule unattended.

Environment

  • Claude Code version: latest
  • Platform: macOS
  • Trigger type: Scheduled remote trigger (CCR)
  • Connectors: Gmail, Google Calendar, Granola, Atlassian (Jira)

## Steps to Reproduce

  1. Connect MCP connectors at claude.ai/settings/connectors
  2. Create a scheduled trigger via RemoteTrigger API with those connectors

attached in mcp_connections

  1. Manually run the trigger (action: "run") — MCP tools load successfully
  2. Wait for the cron schedule to fire automatically (unattended)
  3. Check run logs — agent reports "No MCP tools are loaded in this session"

What Should Happen?

Expected Behavior

MCP connector tools (Gmail, Calendar, etc.) should load in both manual and
scheduled runs since the same connector UUIDs are used in both cases.

Error Messages/Logs

## Actual Behavior
  Manual/test runs: MCP tools load correctly, agent can call Gmail, Calendar etc.
  Scheduled (cron) runs: "No MCP tools are loaded in this session" — agent sees
  only built-in tools (Bash, Read, Write, etc.)

  ## Hypothesis
  The connector OAuth tokens appear to be scoped to interactive/browser sessions.
  Manual runs piggyback an active user session for token exchange. Scheduled
  unattended runs have no live session and need stored offline refresh tokens —
  these either aren't being stored or weren't granted offline_access scope during
  the connector setup flow.

  ## Additional Observation
  After disconnecting and reconnecting connectors, the schedule skill API
  consistently returns "No connected MCP connectors found" even though connectors
  are visible and active in claude.ai/settings/connectors — suggesting a sync
  lag between the UI and the CCR API layer.

  ## Trigger ID (for investigation)
  trig_01KWr7GFhTDZCopuHZXZ6G8i

Steps to Reproduce

## Steps to Reproduce

  1. Connect MCP connectors at claude.ai/settings/connectors
  2. Create a scheduled trigger via RemoteTrigger API with those connectors

attached in mcp_connections

  1. Manually run the trigger (action: "run") — MCP tools load successfully
  2. Wait for the cron schedule to fire automatically (unattended)
  3. Check run logs — agent reports "No MCP tools are loaded in this session"

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.89

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_

View original on GitHub ↗

6 Comments

bart-stealthium · 6 months ago

This is definitely a regression – my scheduled runs worked yesterday around 8:00 UTC and are completely failing today. I'm unable to get it working, even though MCPs are enabled in the scheduled run configuration. Team, please investigate. Cc @ashwin-ant

adamyonk · 6 months ago

Adding another data point. I have two scheduled triggers using Fireflies + Slack MCP connectors (and one also using GitHub MCP).

Working run: session_01CDpFRGLAf1c92nzdVPHhGd — MCP tools loaded immediately on first ToolSearch call. Agent completed successfully. This was the first manual test run after creating the trigger.

Failing runs: Every subsequent run (both scheduled and manual "Run Now") on 2026-04-01 fails. ToolSearch returns "No matching deferred tools found" even with:

  • Exact select: queries (e.g. select:mcp__Fireflies__fireflies_get_transcripts,mcp__Slack__slack_send_message)
  • Sleep+retry loops (up to 5 retries with 10s sleep between each)

In one failing run, the first ToolSearch returned: "No matching deferred tools found. Some MCP servers are still connecting: Fireflies. Their tools will become available shortly — try searching again." — but subsequent retries still failed, suggesting the servers never finish connecting.

Trigger IDs: trig_01SkhDvVUAPBoiFxak6F4Nxk, trig_01F2eYDycfLbVtg2CHpv1tVt
Environment: env_011HaCCkKSGZZ5SumrAM5MRh (anthropic_cloud)
MCP connectors: Fireflies (api.fireflies.ai/mcp), Slack (mcp.slack.com/mcp), GitHub (api.githubcopilot.com/mcp)

ashwin-ant collaborator · 6 months ago

Fix incoming for this

mark-7t-world · 6 months ago

Same issue. Three triggers using Gmail, Slack, Calendar, and Granola MCP connectors. Last successful execution: April 1. Since then, triggers fire (updated_at advances) but sessions fail before any MCP tool calls execute. Manual run via API also fails. The /scheduled UI now shows "Failed to load scheduled tasks." Environment: env_016DzjWDE3TvK7TTdym3BDkm.

ashwin-ant collaborator · 6 months ago

Duplicate of #35899.

github-actions[bot] · 6 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.