[BUG] Deep Research spawned 105 agents and exhausted Max plan quota in ~10 minutes

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 Code Deep Research unexpectedly spawned approximately 105 agents and exhausted the entire quota of my Claude Max plan in less than 10 minutes.

The issue was triggered by a relatively simple research request and the amount of generated agents and consumed tokens appears disproportionate.

What Should Happen?

Claude Code should have performed a reasonably scoped Deep Research workflow for this request.

A prompt asking whether Gemma 4 12B is suitable for a local deployment should not automatically spawn more than 100 agents.

Claude should:

• limit agent fan-out to a reasonable number;
• warn users before launching very large research jobs;
• provide an estimate of expected token usage and agent count;
• prevent runaway agent spawning that can unexpectedly exhaust plan quotas.

Error Messages/Logs

Steps to Reproduce

1. Open Claude Code.
2. Start a Deep Research workflow.
3. Submit the following prompt:

"voglio prima rivalutare ancora altri modelli che possano veramente girare in locale, per esempio gemma 4 12B potrebbe fare la caso nostro?"

1. Observe that Deep Research spawns a very large number of agents (approximately 105/106).
2. The workflow consumes approximately 4.8M tokens and exhausts the Max plan quota.

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

2.1.158 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

_No response_