[BUG] Claude.ai MCPs not showing up on Claude code
Status Closed — not planned
Reported on v2.1.25
Maintainer reply None cached
Workaround ✓ Mentioned in thread ↓
Activity 11 comments · opened Jan 30, 2026 · closed Apr 26, 2026
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?
Previously, I was able to access all my Claude.ai MCPs in Claude code. Recently, I notice that it is missing.
I suspect it is because the flag ("tengu_claudeai_mcp_connectors") keeps being set to false on start up of Claude code. Not sure why this is happening
What Should Happen?
It should always load the Claude.ai MCPs in my Claude code
Error Messages/Logs
Steps to Reproduce
- Set "tengu_claudeai_mcp_connectors": false as true in .claude.json in user scope
- restart claude code
- It will be reset back to false and the MCPs are missing
Claude Model
Opus
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.25
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
VS Code integrated terminal
Additional Information
_No response_
11 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
👎 Not similar to 2 and 3 at all. It looks similar to 1 but the OS and how issue happens seems different.
It would be great to know when this would be brought back, as this brings a critical missing piece to my daily workflow (Krisp MCP via claude.ai).
Additional investigation findings
We investigated this issue across two macOS machines with the same Claude Max account, same organization, and same Claude Code version (2.1.37).
Environment
| | Machine A (working) | Machine B (not working) |
|---|---|---|
|
tengu_claudeai_mcp_connectors|true|false(reset on every restart) || Claude Code version | 2.1.37 | 2.1.37 |
| Account/email | Same | Same |
| Organization UUID | Same | Same |
What we ruled out
checkGateCached():DISABLE_TELEMETRY— not set on either machineCLAUDE_CODE_USE_BEDROCK— not setCLAUDE_CODE_USE_VERTEX— not setCLAUDE_CODE_USE_FOUNDRY— not setCLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC— not set.zshrc,.bashrc,.zprofile,.bash_profile,.zshenv) — no matcheslaunchctl getenv— not setanonymousId— Different between machines (claudecode.v1.{UUID}). Copying the working machine'sanonymousIdto the non-working machine did not change the flag evaluation result.oauthAccountfields — The non-working machine was missingorganizationRole,workspaceRole, andorganizationName. Manually adding them did not help.tengu_claudeai_mcp_connectors: truein~/.claude.jsonworks within the running session, but gets overwritten tofalseon the next startup by server sync.claude logout→claude login, reconnecting Slack OAuth from claude.ai web — none helped.Conclusion
The issue appears to be purely server-side feature flag evaluation (GrowthBook). Same account, same org, same version, same env — but different
anonymousIdvalues get different gate results. No client-side workaround found.Workaround: SessionStart hook to force the flag on every launch
Since the flag gets overwritten on every restart by server-side sync, I set up a Claude Code
SessionStarthook that forcestengu_claudeai_mcp_connectorsback totruebefore the session begins.~/.claude/scripts/fix-mcp-gate.sh~/.claude/settings.json(add tohooks.SessionStart):Requires
jq(brew install jq). The hook runs before the session starts, so MCP connectors are available immediately.Not a real fix, but it keeps things working until the server-side flag evaluation is resolved.
the SessionStart hook ~~didn't work for me~~. But this script did, after modifying it to target the claudeai_mcp_connectors feature flag: https://gist.github.com/gastonmorixe/9c596b6de1095b6bd3b746ca3a1fd3d7
EDIT: Actually, the SessionStart hook DID work after a reboot! Thanks!
Please, don't make me trust an insecure third-party MCP server with my email!
Please, either fix this, or provide an authorized/"officially from Anthropic" MCP server for GMail integration.
This is still an active issue. Above fix did nothing for me. Relevant logs:
``
2026-03-03T18:34:27.545Z [DEBUG] [claudeai-mcp] Fetch failed``2026-03-03T18:34:27.552Z [ERROR] AxiosError: [url=https://platform.claude.com/v1/oauth/token,status=400] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
2026-03-03T18:34:27.124Z [DEBUG] [claudeai-mcp] Checking gate (cached)...
2026-03-03T18:34:27.124Z [DEBUG] [claudeai-mcp] Gate returned: true
2026-03-03T18:34:27.125Z [DEBUG] [claudeai-mcp] Fetching from https://api.anthropic.com/v1/mcp_servers?limit=1000
2026-03-03T18:34:27.125Z [DEBUG] Found 1 enabled plugins
2026-03-03T18:34:27.126Z [DEBUG] All plugins already exist, skipping migration
2026-03-03T18:34:27.126Z [DEBUG] Found 3 installed plugins
2026-03-03T18:34:27.126Z [DEBUG] Found 0 missing plugins (not installed):
2026-03-03T18:34:27.126Z [DEBUG] Setting installation status: 0 marketplaces, 0 installable plugins, 0 uninstallable plugins
2026-03-03T18:34:27.127Z [ERROR] AxiosError: [url=https://api.anthropic.com/api/oauth/claude_cli/client_data,status=429,body=Rate limited. Please try again later.] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
2026-03-03T18:34:27.127Z [ERROR] AxiosError: [url=https://api.anthropic.com/api/oauth/claude_cli/client_data,status=429,body=Rate limited. Please try again later.] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
2026-03-03T18:34:27.128Z [ERROR] AxiosError: [url=https://api.anthropic.com/api/oauth/claude_cli/client_data,status=429,body=Rate limited. Please try again later.] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
2026-03-03T18:34:27.128Z [ERROR] AxiosError: [url=https://api.anthropic.com/api/oauth/claude_cli/client_data,status=429,body=Rate limited. Please try again later.] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
2026-03-03T18:34:27.128Z [ERROR] AxiosError: [url=https://api.anthropic.com/api/oauth/claude_cli/client_data,status=429,body=Rate limited. Please try again later.] Error
at YM (/$bunfs/root/claude:92:1144)
at <anonymous> (/$bunfs/root/claude:93:10067)
at emit (node:events:92:22)
at endReadableNT (internal:streams/readable:861:50)
at processTicksAndRejections (native:7:39)
at request (/$bunfs/root/claude:95:2149)
at processTicksAndRejections (native:7:39)
Please fix this bug
Closing for now — inactive for too long. Please open a new issue if this is still relevant.
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.