Summary

#26824 (salvage of #10273, merged 2026-05-16) added a new delegation.api_mode config key so subagent delegation can honor an explicit wire-protocol override when _detect_api_mode_for_url() can't classify a custom base_url (Azure AI Foundry, MiniMax, Zhipu GLM, LiteLLM proxies, …). hermes_cli/config.py documents the key inline, but the user-facing website docs still only mention model, provider, base_url, and api_key for the direct-endpoint path, so users hitting the 404 #26824 fixes have no docs to point at.

This patch surfaces api_mode in the two user-guide pages that already enumerate the delegation YAML keys.

Changes

• website/docs/user-guide/configuration.md (Delegation section): add the # api_mode: "" line to the commented YAML example next to api_key, and a new Wire protocol (api_mode) paragraph after Direct endpoint override that paraphrases the auto-detect rules and lists the three valid values.
• website/docs/user-guide/features/delegation.md (Configuration subsection): add a commented api_mode line to the "Or use a direct custom endpoint" YAML block so the example matches the one in configuration.md.

Value set (chat_completions, codex_responses, anthropic_messages) and auto-detect behavior are taken verbatim from the hermes_cli/config.py docstring written by #26824. No claim is made about auto-detect handling endpoints beyond the URL/known-host heuristics in _detect_api_mode_for_url().

4 LOC additive total; no source-code change.

Test plan

• Documentation-only change.
• Verified the value set against hermes_cli/config.py lines 1149-1152 (added by fix(delegation): honor api_mode + auto-detect anthropic_messages URLs #26824).
• Verified auto-detect rules against fix(delegation): honor api_mode + auto-detect anthropic_messages URLs #26824's PR description: /anthropic suffix → anthropic_messages, plus retained chatgpt.com/backend-api/codex, api.anthropic.com, api.kimi.com/coding hostname overrides.
• Sibling-PR poach guard: no OPEN PR touches website/docs/user-guide/configuration.md Delegation section or features/delegation.md for api_mode (docs: document explicit api_mode for custom endpoints #7176 covers model.api_mode in integrations/providers.md, different config namespace).