Follow Discord
Sweep 22 Oct 2026 · 21:04Z Build v2.1.309 528 read Stable v2.1.308 Latest v2.1.309 Next v2.1.309 Feeds RSS JSON llms.txt llms-full.txt Unofficial

Claude Code v2.1.304 ·

CLAUDE_CODE_AUTO_COMPACT_WINDOW gets shorthand values, warnings and a note in /config

The auto-compact window env var accepts 200k or 1M shorthand, clamps or warns on bad values, and /config shows an extra note about it

Group of 3 Use it now Improvements
JSON All of v2.1.304
Use it nowTier: how much it should matter to you
3Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
CompactionArea: what it touches
ImprovementsKind: in v2.1.304,
What probably matters to youSection of the release

Unclear What the warning says is not stated.

What

Auto-compact is when Claude Code summarises the conversation so far to free up room in the model's memory. The auto-compact window sets how full that memory can get, in tokens, before this happens. CLAUDE_CODE_AUTO_COMPACT_WINDOW is an environment variable, a value set in your shell, that sets this window. This release changes how it is read and shown:

  • The value is read by a new parser. It accepts shorthand such as 200k or 1M, and a bare number between 100 and 1000 is treated as thousands.
  • Values that are too small or too large are brought up to the minimum or down to the maximum, with a warning in the log or in headless output.
  • A value that cannot be read at all is ignored, with the message that it "isn't a usable size, so Claude Code ignores it." Before, an invalid value was dropped without a visible warning.
  • At startup in interactive mode, the variable is checked and a warning is raised if needed.
  • The auto-compact window entry in /config now shows an extra warning-coloured line built from the environment variable. Its exact wording was not read. Before, the only extra line was "Auto-compact is currently disabled (see /config)".

Why

A mistyped window size now produces a warning instead of quietly having no effect, and the settings screen points out when the environment variable is in play.

Read from
Names in the bundle/config
Since it was published

The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.

Confirmed since Anthropic's documentation has since written up /config, on Connect to external tools with MCP. To choose the first-turn wait yourself, set `CLAUDE_CODE_MCP_STARTUP_WAIT_MS` in the [`env` option](/docs/en/agent-sdk/configuration#set-environment-variables), for example `CLAUDE_CODE_MCP_STARTUP_WAIT_MS: "5000"`. The first turn then wai… agent-sdk/mcp see the edit
Added since A small documentation edit on Use Claude Science on a corporate network touched a line naming /config after this was published. The `[network] ca_bundle` key in `config.toml` points at a PEM file whose certificates Claude Science adds to its default trust for sign-in, the Claude API, Anthropic-hosted connectors, update checks, and [cloud storage](/docs/claude-scien… claude-science/corporate-networks see the edit
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtWhat the warning says is not stated.
Anthropic's documentation agreesA small documentation edit on Use Claude Science on a corporate network touched a line naming /config after this was published.
Anthropic's release notes agreeFixed the organization policy not refreshing until the hourly check after approving an ANTHROPIC_API_KEY at startup or switching "Use custom…

See this entry in the whole of v2.1.304 →

Feedback