Product feature · runtime
Subagent resultsCommon product term
Return a child agent's findings or work to the parent.
Terminology basis: Common product term. xAI — Grok Bot messaging and collaboration
Subagent results: 8 supported, 0 partial, 0 unsupported, 23 unreviewed across 31 cataloged products.
Current evidence by product
Can my agent use Subagent results?
Read across for the answer. 8 of 31 current product columns have reviewed evidence; unreviewed does not mean unsupported.
Web
9 productsDesktop
13 productsCLI
9 productsDefinition and scope
What this capability means
This row asks what a child agent returns to its parent. A plain text summary counts as a result; files, patches, citations, status, tool output, and structured objects are additional result forms recorded as qualifiers.
Evidence should identify documented result types, size limits, streaming versus final delivery, partial results after cancellation, error representation, and whether created files remain available outside the child’s workspace. A typed result envelope is not required for support.
Traceable compatibility
Assertion ledger
Documentation evidence only. No runtime conformance test is implied.
- Target
- 2026-08-28 ChatGPT Work web documentation observation · hosted-observation
- Environment
- hosted-default
- Observed
- 2026-08-28
- runtimeactivity, results, and child threads are inspectable and consolidated into the response, but no typed artifact, citation, diff, status, or error schema is documented
- OpenAI — Subagents in ChatGPT and Codexdocumented · reviewed 2026-08-28
- Target
- current ChatGPT desktop documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtimechild threads and summaries remain inspectable, but no typed artifact, citation, diff, status, or error schema is documented
- OpenAI — Subagents in ChatGPT and Codexdocumented · reviewed 2026-08-28
- Target
- current Codex CLI documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtime/agent exposes child threads and the parent collects results, but no typed artifact, citation, diff, status, or error schema is documented
- OpenAI — Subagents in ChatGPT and Codexdocumented · reviewed 2026-08-28
- Target
- current Gemini CLI documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtimefindings return to the parent, but no typed artifact, citation, diff, status, or error schema is documented
- Google — Gemini CLI subagentsdocumented · reviewed 2026-08-28
- Target
- current Cursor subagent documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtimea final message and child branch changes remain available, but no typed artifact, citation, status, or error schema is documented
- Cursor — Subagentsdocumented · reviewed 2026-08-28
- Target
- current VS Code subagent documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtimethe parent receives a final result and the operator can inspect the child trace, but no typed artifact, citation, diff, status, or error schema is documented
- Visual Studio Code — Subagents in Visual Studio Codedocumented · reviewed 2026-08-28
- Target
- current Claude Code subagent documentation · dated-documentation
- Environment
- local-default
- Observed
- 2026-08-28
- runtimethe parent receives a prose summary and can observe child file changes, but no typed result schema for citations, artifact identities, status, or errors is documented
- Anthropic — Create custom subagentsdocumented · reviewed 2026-08-28
- Target
- 2026-08-28 Grok Bot desktop documentation observation · hosted-observation
- Environment
- hosted-default
- Observed
- 2026-08-28
- runtimevisible asynchronous messages and shared files preserve useful handoff context, but Bot-to-group handoff messages are text-only and no typed result schema is documented
- policyshared files live on one account-scoped computer and do not create an isolated provenance boundary
- xAI — Grok Bot messaging and collaborationdocumented · reviewed 2026-08-28
- 1. Evidence checked 2026-08-28: Grok Bots can hand work to one another through visible asynchronous messages and shared files. Bot-to-group handoff messages are text-only, while images can be sent directly to another Bot; the docs do not define a typed result envelope for status, citations, or errors.
- 2. Evidence checked 2026-08-28: Claude Code subagents report a final summary to the parent conversation and can preserve file edits in the shared workspace or an isolated worktree. The docs do not define a typed result envelope for citations, status, artifacts, or errors.
- 3. Evidence checked 2026-08-28: ChatGPT Work and Codex collect child results into the parent response and preserve inspectable child threads, but the reviewed documentation does not define a typed citation, artifact, diff, or error envelope.
- 4. Evidence checked 2026-08-28: Gemini CLI subagents report findings back to the main agent after completing an isolated context loop; no typed result envelope is defined on the reviewed page.
- 5. Evidence checked 2026-08-28: Cursor subagents return a final message to the parent and can leave work on shared or isolated branches; the reviewed page does not define typed result provenance.
- 6. Evidence checked 2026-08-28: VS Code returns only a child's final result to the parent while retaining the prompt, full tool trace, and returned result in an inspectable read-only child chat; no typed result envelope is documented.
No sourced issues are attached to this row.
- Methodology note
- xAI — Grok Bot messaging and collaboration documented · xAI · reviewed 2026-08-28
- Anthropic — Create custom subagents documented · Anthropic · reviewed 2026-08-28
- OpenAI — Subagents in ChatGPT and Codex documented · OpenAI · reviewed 2026-08-28
- Google — Gemini CLI subagents documented · Google · reviewed 2026-08-28
- Cursor — Subagents documented · Cursor · reviewed 2026-08-28
- Visual Studio Code — Subagents in Visual Studio Code documented · Microsoft · reviewed 2026-08-28
Found a wrong or incomplete result?
Name the exact product, describe what happened, add the date, and link public evidence when available. A community report starts a review. It does not change the support state by itself.