[BUG] SendMessage tool referenced in Agent tool documentation but not available at runtime

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?

The Agent tool documentation states:

▎ "To continue a previously spawned agent, use SendMessage with the agent's ID or name as the to field — that resumes it with full context. A new Agent call starts a
▎ fresh agent with no memory of prior runs."

However, SendMessage is not available as a tool at runtime — neither as a direct tool nor as a deferred tool loadable via ToolSearch.

Environment

• Claude Code version: 2.1.104
• Model: claude-opus-4-6
• Platform: Linux (WSL2) — Linux 6.6.87.2-microsoft-standard-WSL2
• Interface: CLI

What Should Happen?

Expected Behavior

SendMessage should be available as a tool to resume a previously spawned agent, preserving its full conversation context as documented.

Actual Behavior

SendMessage is not available. The only way to interact with agents is Agent(), which always starts fresh. The agent's accumulated context (file reads, grep results,
analysis) is effectively lost after it delivers its report.

Error Messages/Logs

Query: "select:SendMessage"                                                                                                                                            
  Result: "No matching deferred tools found"

Steps to Reproduce

Steps to Reproduce

1. Spawn a background agent using Agent() — it completes and returns an agentId
2. Attempt to continue the conversation with that agent using SendMessage
3. SendMessage does not exist in the available toolset
4. Using ToolSearch with query select:SendMessage returns: "No matching deferred tools found"
5. The only option is Agent() which creates a new agent with no context from the previous one

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.104

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

Environment

• Claude Code version: 2.1.104
• Model: claude-opus-4-6
• Platform: Linux (WSL2) — Linux 6.6.87.2-microsoft-standard-WSL2
• Interface: CLI

Impact

• Tokens wasted when needing to ask follow-up questions to an agent (must re-explore from scratch)
• The documented workflow of spawning an agent, getting results, then asking follow-ups is broken
• Misleading documentation creates false expectations about agent lifecycle