What changed
A background subagent is a second agent that keeps working while you stay in the main Claude Code chat. Its work still counts toward your usage limit. [ClaudeDevs](https://x.com/ClaudeDevs/status/2118063143167693011) says a running subagent updates its status line every 30 seconds, and that each update used to re-read its whole conversation. ClaudeDevs says Haiku 5.5 now writes that line from short notes, and that each update costs about 99 percent less. The post does not say what those short notes contain. ClaudeDevs says the subagent itself keeps its own model. The 99 percent figure in the post is about each status update.
ClaudeDevs says sessions with background subagents used about 5 percent less of the limit, so the same limit gets you about 6 percent more prompts. The post gives the 5 and 6 percent figures in one sentence, without a sample size, a usage window, or the arithmetic that links them.
Who gets it
ClaudeDevs says the change is live for most Pro, Max, and Ultra users and for everyone on Team and Enterprise. The post does not say which of those Pro, Max, and Ultra users count as most. Changelog notes for v2.1.316, v2.1.317, and v2.1.318 do not list this eligibility. The post states no dollar price, token rate, or new plan cap.
How to try it
The only instruction in the post is to update to v2.1.316 or later. ClaudeDevs names no flag, setting, or command. The changelog says to run claude --version to check the installed version.
The [Claude Code changelog](https://code.claude.com/docs/en/changelog.md) dates v2.1.316 to October 31, 2026, v2.1.317 to November 3, 2026, and v2.1.318 to November 4, 2026, the same calendar day as the ClaudeDevs post at 19:29 UTC.
Other subagent lines in those notes sit beside the announcement. v2.1.316 adds CLAUDE_CODE_SUBAGENT_EFFORT to set the effort level subagents run at when nothing else names one, instead of inheriting the session's. v2.1.317 changes subagents and teammates on a different model from the main conversation to use that model's saved effort level or its default, instead of the session's effort level. v2.1.318 moves a background subagent's shell command to the background after 3 minutes on all providers, so the prompt cache does not expire while it runs. On Bedrock, Vertex, Foundry, or with telemetry off, a subagent progress summary on a cold prompt cache can wait up to 5 minutes for the subagent's pending response. v2.1.318 also fixes the "Now using usage credits" notice reappearing about every 30 seconds while a background subagent is running. That 30-second interval is the notice. ClaudeDevs' 30-second interval is the status refresh that used to re-read the conversation. The effort setting, the shell move, the cold-cache wait, and that notice are separate changelog items. The Haiku status writer appears in the ClaudeDevs post.
What is still unclear
The name Haiku 5.5 already appears in v2.1.311, dated October 26, 2026. In that note, ultrathink in the interactive terminal runs at xhigh effort on Opus 5.5, Sonnet 5.5, Haiku 5.5, and Fable 5.1 through Anthropic's API. That sentence is about effort on a turn. The 99, 5, and 6 percent figures are the post's. The post does not say what the short notes contain. Those version dates do not show when the status-line change shipped.
Source
Also from
Sources and editorial approach
This article relies on the ClaudeDevs post published November 4, 2026, at 19:29 UTC (https://x.com/ClaudeDevs/status/2118063143167693011) and the Claude Code changelog (https://code.claude.com/docs/en/changelog.md), checked November 4, 2026. The changelog sentences draw on the notes for v2.1.318 (November 4, 2026), v2.1.317 (November 3, 2026), v2.1.316 (October 31, 2026), and v2.1.311 (October 26, 2026). Older entries further down the page are not part of this article. The changelog page says it is generated from the GitHub CHANGELOG.md, and that GitHub file is not covered here. The percentage claims were not checked against a usage session.