Anthropic's `claude-code-darwin-arm64` binary update includes new internal API endpoints a

EXHIBIT 817 · URGENT

Aug 19, 2026 00:48 UTC binary @anthropic-ai/claude-code-darwin-arm64

Anthropic's `claude-code-darwin-arm64` binary update includes new internal API endpoints and mentions "experimental.evals" features, suggesting potential new evaluation tooling or capabilities.

READY TO POST:

anthropic's latest claude-code binary has some new endpoints, including references to "experimental.evals." could we see new testing or evaluation features coming soon?

just spotted an update to anthropic's claude-code binary with new internal api paths and mentions of 'experimental.evals'. looks like they might be building out more developer tooling.

⚠️ FROM COMPILED BINARY — lower confidence, verify before posting. (May include bundled-dependency strings, not just first-party changes.) @anthropic-ai/claude-code-darwin-arm64 2.1.234 -> 2.1.235: +7293 strings, -6038 strings, 20 interesting [feature_flag] (8) + <${$H}>${n.length} background ${r} task(s) from the previous session have no completion record. ${l} They have been marked ${e}. ${c} Task ids in this notification beginning with "${Lms}" are internal scan markers, not tasks.</${$H}> + Directory (below the current directory) to write cases into (default: experimental.evals from the plugin.json in the current directory, else evals/) + experimental.evals + experimental.evals in + ignoring experimental.evals + ignoring experimental.evals in + internal error: + the manifest's experimental.evals [endpoint] (12) + /api/frame/contract/latest + /api/oauth/organizations/:orgUUID/mcp/connectors/listc + /v1/toolbox/shttp/mcp/{server_id}Wa + <r>/M1fdb5e06cce6b6c535b58f061171c83311a0af86a/viewhttps://bun.report + http://www.ibm.com/data/dtd/v11/ibmxhtml1-transitional.dtd! + https://aiplatform. + https://aiplatform.googleapis.comSP + https://claude.ai/chrome + https://claude.ai/restricted + https://claude.ai/settings/usage?from=cc_cli_limit_message! + https://github.com/anthropics/claude-code-action/blob/main/docs/setup.md + https://status.claude.comt [sentence] (2) + <${$H}>${n.length} background ${r} task(s) from the previous session have no completion record. ${l} They have been marked ${e}. ${c} Task ids in this notification beginning with "${Lms}" are internal scan markers, not tasks.</${$H}> + Directory (below the current directory) to write cases into (default: experimental.evals from the plugin.json in the current directory, else evals/)