[BUG] Claude Desktop blocks CLI-native SendMessage, breaking subagent continuation

Status Open
Reported on v2.1.260
Maintainer reply None cached
Activity 0 comments · opened Sep 5, 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?

In Claude Desktop's Code tab, SendMessage is missing from both the direct and the deferred tool lists (ToolSearch "select:SendMessage" returns "No matching deferred tools found"), and the app denies the CLI-native tool before it runs. SendMessage is the only way to continue an existing subagent with its context. The desktop replacement takes a session_id and cannot address a subagent, so a rule written for unreachable cross-session targets removed the one target that IS reachable: this session's own subagent. Also note the asymmetry: I can message a session I have no tool to create, but cannot message the only thing I can create.

What Should Happen?

SendMessage should stay available in desktop sessions when the target is an in-process subagent of that session, so review rounds can continue the same subagent instead of starting a fresh one.

Error Messages/Logs

PreToolUse deny, telemetry reason: cli_native_send_message
"Use <desktop tool> to message another session from the desktop app; the CLI-native SendMessage tool is unavailable here (terminal sessions outside Claude Desktop are not reachable from this session)."

Steps to Reproduce

  1. Open a session in Claude Desktop's Code tab.
  2. Spawn a subagent with the Agent tool and let it finish.
  3. Run ToolSearch with query "select:SendMessage" -> "No matching deferred tools found".
  4. ListAgents is present and its description still instructs addressing agents via SendMessage.
  5. Run the same bundled runtime from a terminal (%APPDATA%\Claude\claude-code\2.1.260\claude.exe) and repeat step 3 -> the tool IS listed.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Worked in desktop sessions on the app build carrying Claude Code 2.1.258 (Sept 3 - Sept 4). Broke with the app build installed Sept 4.

Claude Code Version

2.1.260

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

Not a terminal session: Claude Desktop 1.46388.4.0 (Windows 11, MSIX install), Code tab, entrypoint claude-desktop. The app pins Claude Code 2.1.260 and its runtime cannot be updated independently, so I cannot test 2.1.261 from the app.

Cost per review round, same task: continuing an existing subagent 13 min / 34 file reads; a fresh subagent 23 min / 69 reads, with repeated findings.

View original on GitHub ↗