Issue Station
Sign In
Tools hereToolbox

claude-code - 💡(How to fix) Fix [BUG] MCP tools unavailable in agent-team exec subagents — mcp__*__* returns "No such tool available" despite servers being connected at session level [1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#52669•Fetched 2026-04-24 10:42:55
View on GitHub
Comments
0
Participants
1
Timeline
5
Reactions
0
Author
DougLifeOfficialDougLifeOfficial
Participants
DougLifeOfficialDougLifeOfficial
Timeline (top)
labeled ×5

When using the experimental Agent Teams feature (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1), any mcp__*__* tool call made inside an exec subagent process returns No such tool available. The MCP servers are configured and connected at the parent session level (system-layer info is injected, /mcp reports them as connected), but the tool registry does not appear to be exposed to the exec subagent spawned by the agent-teams runtime.

This affects every MCP server configured in the session — it is not specific to any one server — so it appears to be a generic tool-registry propagation issue rather than a per-server connection problem.

The main agent can call the mcp_tools but none of the subagents can.

Error Message

Error Messages/Logs

  1. Observe: the call returns No such tool available. Every mcp__*__* tool for every configured MCP server fails with the same error.

Root Cause

When using the experimental Agent Teams feature (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1), any mcp__*__* tool call made inside an exec subagent process returns No such tool available. The MCP servers are configured and connected at the parent session level (system-layer info is injected, /mcp reports them as connected), but the tool registry does not appear to be exposed to the exec subagent spawned by the agent-teams runtime.

This affects every MCP server configured in the session — it is not specific to any one server — so it appears to be a generic tool-registry propagation issue rather than a per-server connection problem.

The main agent can call the mcp_tools but none of the subagents can.

Code Example



---

export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

---

claude
RAW_BUFFERClick to expand / collapse

Preflight Checklist

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

What's Wrong?

Summary

When using the experimental Agent Teams feature (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1), any mcp__*__* tool call made inside an exec subagent process returns No such tool available. The MCP servers are configured and connected at the parent session level (system-layer info is injected, /mcp reports them as connected), but the tool registry does not appear to be exposed to the exec subagent spawned by the agent-teams runtime.

This affects every MCP server configured in the session — it is not specific to any one server — so it appears to be a generic tool-registry propagation issue rather than a per-server connection problem.

The main agent can call the mcp_tools but none of the subagents can.

Minimal reproduction

Prereqs: at least one MCP server already configured and connected (any server will do — confirm with /mcp).

  1. Enable agent teams: export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1.
  2. Launch Claude Code (claude) in the project.
  3. From the lead session, spawn a teammate (e.g. via the Task/agent-teams entry point) so the exec subagent process boots.
  4. Ask the teammate, on its first turn, to call any mcp__<server>__<tool> that works in the parent session.
  5. The call returns No such tool available.

Expected: the MCP tool executes just like it does in the parent session. Actual: No such tool available for every mcp__*__* tool, for every configured MCP server.

Expected behavior

Exec subagents spawned under agent teams should inherit the parent session's MCP tool registry (or synchronously load it before the first turn), so mcp__*__* tools are callable on turn 1 just as they are in the parent session.

Actual behavior

MCP tools are absent from the subagent's tool registry and every mcp__*__* call fails with No such tool available.

What I've verified

  • The MCP servers are configured correctly.
  • They are connected at the session level (system-layer info is injected into the parent session; /mcp shows them as connected).
  • The failure is reproducible across all configured MCP servers, not isolated to one.

Related / prior reports

This looks like the same underlying problem tracked in #23625 (auto-closed as stale), with additional context in #25200 and a proposed fix direction in #31623. Filing fresh per the stale-closure guidance, since the issue is still reproducible on current Claude Code.

Environment

  • Claude Code version: 2.1.119
  • OS: Windows (running Claude Code under WSL + tmux)
  • Feature flag: CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

What Should Happen?

Exec subagents spawned via the agent-teams runtime should have the parent session's MCP tools available on their first turn. mcp__<server>__<tool> calls should succeed inside the subagent just as they do in the parent session, instead of returning "No such tool available".

Error Messages/Logs

Steps to Reproduce

Prereqs: at least one MCP server configured and connected at the session level (any server will do — confirm with /mcp before starting).

  1. Enable agent teams: 
    export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
  2. Launch Claude Code in a project: 
    claude
  3. In the lead session, confirm MCP servers are connected (/mcp lists them as connected and the same mcp__<server>__<tool> calls work from the lead).
  4. Spawn a teammate / exec subagent via the agent-teams entry point and let it boot.
  5. On the teammate's first turn, ask it to call any mcp__<server>__<tool> that is known to work in the lead.
  6. Observe: the call returns No such tool available. Every mcp__*__* tool for every configured MCP server fails with the same error.

Expected: the MCP tool executes just like it does in the parent/lead session. Actual: every mcp__*__* call in the subagent returns No such tool available, even though the servers are connected at the session level.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

2.1.119

Platform

Anthropic API

Operating System

windows

Terminal/Shell

WSL/UBUNTU w/ TMUX

Additional Information

No response

extent analysis

TL;DR

The issue can be resolved by ensuring the MCP tool registry is properly propagated to the exec subagent process spawned by the agent-teams runtime.

Guidance

  • Verify that the MCP servers are correctly configured and connected at the session level by checking the output of /mcp in the parent session.
  • Investigate the tool registry propagation mechanism in the agent-teams runtime to ensure it is correctly exposing the MCP tools to the exec subagent process.
  • Check the related issues (#23625, #25200, and #31623) for potential fixes or workarounds that may apply to this scenario.
  • Consider testing the issue with different MCP servers or tool configurations to determine if the problem is specific to a particular setup.

Example

No code snippet is provided as the issue does not contain sufficient information to create a specific example.

Notes

The issue appears to be related to a regression in the Claude Code version 2.1.119, and the fix may involve updating the tool registry propagation mechanism in the agent-teams runtime.

Recommendation

Apply a workaround by investigating the tool registry propagation mechanism and ensuring it is correctly exposing the MCP tools to the exec subagent process, as the root cause of the issue is likely related to this mechanism.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

Exec subagents spawned under agent teams should inherit the parent session's MCP tool registry (or synchronously load it before the first turn), so mcp__*__* tools are callable on turn 1 just as they are in the parent session.

#api#configuration error#environment variable#network issue#logging issue

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

claude-code - 💡(How to fix) Fix [BUG] MCP tools unavailable in agent-team exec subagents — mcp__*__* returns "No such tool available" despite servers being connected at session level [1 participants]