Answer: Set the `effort` option in the top-level `query()` options for the session, or per subagent with the `effort` field on `AgentDefinition` to override the session level. The levels are "low", "medium", "high", "xhigh", and "max". Status: Confirmed Entry: tool/claude-code/agents-and-automation/set-effort-level Revision: 2 Digest: sha256:f116bda1a08babb7 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:02:36Z Last changed: 2026-11-18T12:12:10Z (where to set (subagent): — → per subagent with the `effort` field on [`AgentDefinition`]) Last verified: 2026-11-20T12:16:16Z Next check: 2026-11-21T12:16:16Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of option name | effort | [1] | 2026-11-20 where to set (subagent) | per subagent with the `effort` field on [`AgentDefinition`] | [1] | 2026-11-20 level low | "low" | [1] | 2026-11-20 level medium | "medium" | [1] | 2026-11-20 level high | "high" | [1] | 2026-11-20 level xhigh | "xhigh" | [1] | 2026-11-20 level max | "max" | [1] | 2026-11-20 Sources: [1] Anthropic. "How the agent loop works". https://code.claude.com/docs/en/agent-sdk/agent-loop.md Role: primary. Accessed: 2026-11-20T12:16:16Z Quote: "The `effort` option controls how much reasoning Claude applies." (Supports: option name) Quote: "or per subagent with the `effort` field on [`AgentDefinition`]" (Supports: where to set (subagent)) Quote: "| `"low"` | Minimal reasoning, fast responses | File lookups, listing directories |" (Supports: level low) Quote: "| `"medium"` | Balanced reasoning | Routine edits, standard tasks |" (Supports: level medium) Quote: "| `"high"` | Thorough analysis | Refactors, debugging |" (Supports: level high) Quote: "| `"xhigh"` | Extended reasoning depth |" (Supports: level xhigh) Quote: "| `"max"` | Maximum reasoning depth | Multi-step problems requiring deep analysis |" (Supports: level max) Published: 2026-11-20T02:53:15Z Snapshot SHA-256: 55a463c38057a4f24a8c6ffe8eb4984a5285d0e3eda21f982a81cc193bb4ad5b (captured 2026-11-20T12:16:13Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. Cross-check: not-available at 2026-11-20T12:16:16Z. No independent second source is configured for this entry Last successful check: 2026-11-20T12:16:16Z Last attempt: 2026-11-20T12:16:16Z Notes and limitations: Extracted automatically from a single official source (section "Effort level"). Related: https://latestn.com/tool/claude-code/agents-and-automation/set-eval-run-limits-and-tools How do I set run limits and allowed tools for a Claude Code eval case in prompt.md? https://latestn.com/tool/claude-code/agents-and-automation/set-up-first-hook How do I set up a hook in Claude Code to run a shell command automatically? https://latestn.com/tool/claude-code/agents-and-automation/setting-sources-not-controlled Which settings does settingSources in the Claude Code SDK not control? https://latestn.com/tool/claude-code/agents-and-automation/session-storage-supported-platforms Which platforms and runtimes support external session storage in the Claude Code Agent SDK? https://latestn.com/tool/claude-code/agents-and-automation/sessionstore-interface-methods What methods does the Claude Code SessionStore interface require and which are optional? https://latestn.com/tool/claude-code/agents-and-automation/set-claude-code-preset-system-prompt How do I set the claude_code preset as the system prompt in the Claude Code Agent SDK? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-18T12:12:10Z | changed | where to set (subagent): — → per subagent with the `effort` field on [`AgentDefinition`] | [1] 2026-11-18T12:12:10Z | changed | level max: max → "max" | [1] 2026-11-18T12:12:10Z | changed | level xhigh: xhigh → "xhigh" | [1] 2026-11-18T12:12:10Z | changed | level high: high → "high" | [1] 2026-11-18T12:12:10Z | changed | level medium: medium → "medium" | [1] 2026-11-18T12:12:10Z | changed | level low: low → "low" | [1] 2026-11-13T03:02:36Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/claude-code/agents-and-automation/set-effort-level/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/claude-code/agents-and-automation/set-effort-level POST: https://latestn.com/dispute/tool/claude-code/agents-and-automation/set-effort-level with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/claude-code/agents-and-automation/set-effort-level?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How do I set the effort level for a Claude Code SDK agent? Revision 2, last verified 2026-11-20T12:16:16Z. https://latestn.com/tool/claude-code/agents-and-automation/set-effort-level (accessed 2026-11-21T09:58:57Z). @misc{LatestN_tool_claude_code_agents_and_automation_set_effort_level_r2, title = {How do I set the effort level for a Claude Code SDK agent?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/claude-code/agents-and-automation/set-effort-level}}, note = {Revision 2; digest sha256:f116bda1a08babb7; last verified 2026-11-20T12:16:16Z; accessed 2026-11-21T09:58:57Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/claude-code/agents-and-automation/set-effort-level