creation_account_user
ChatGPTRead the account and available credits.
creation_asset_resources
ChatGPTList candidate resources of one Asset, with default selection and visibility. Continue with next_offset and expected_graph_revision=as_of_seq.
creation_chat_history
ChatGPTRead a bounded history page; validate a restore anchor with restore_preview.
creation_chat_pending_actions
ChatGPTRead only the current valid question or approval of the selected conversation.
creation_composition_get
ChatGPTRead Composition settings, tracks, clips and diagnostics. Clips are paged in track order across all tracks; empty clips on a page do not mean an empty track. Follow next_offset with expected_node_revision=node_revision until has_more=false. This is a partial read view, not a body to submit for replacement.
creation_conversation_branch
ChatGPTBranch at a message into a new conversation in the same project; project_branch instead creates another project.
creation_conversation_create
ChatGPTCreate a separate conversation in a Creation project.
creation_conversation_list
ChatGPTList conversations before choosing which one to continue.
creation_document_search
ChatGPTSearch Document text for bounded excerpts and exact node IDs, lines and content revisions. Narrow node_ids when the project has many documents.
creation_download_all
ChatGPTArchive all visible project media resources and Documents. Reuse client_request_id on retries.
creation_download_status
ChatGPTRead by task_id, or recover a lost submission result with client_request_id. Omit both for the latest archive.
creation_export_readiness
ChatGPTCheck and select a composition before final-video export.
creation_export_status
ChatGPTRead by task_id, or recover a lost submission result with client_request_id. Omit both for the latest export.
creation_export_video
ChatGPTExport the exact composition_node_id and expected_node_revision returned by readiness. Reuse both the revision and client_request_id on retries.
creation_generation_models
ChatGPTDiscover public executable image/video/audio/music models, including member models. Use detail=full for parameters and input roles; membership is enforced when generating.
creation_inspiration_batch_get
ChatGPTRead current published inspiration instructions by ID, up to 50. Unavailable items return per-item errors. This does not authorize or start generation.
creation_inspiration_categor_items
ChatGPTBrowse published inspiration cards by locale and category. Agent recommendations do not filter cards. Follow next_cursor with unchanged filters. Names can match multiple items; never infer uniqueness from a truncated page.
creation_inspiration_categories
ChatGPTList enabled inspiration categories for a locale, including the default category.
creation_inspiration_config
ChatGPTRead configured inspiration tools, Agents and localized installation instructions.
creation_interaction_submit
ChatGPTSubmit the user's explicit response to a server-issued interaction_id. Reuse client_request_id for retries.
creation_project_branch
ChatGPTCopy into a new project at a selected message, with selected/all conversation scope. Use conversation_branch to explore within the same project.
creation_project_create
ChatGPTCreate a Creation project after the goal is clear.
creation_project_document
ChatGPTRead an exact Document node with bounded line ranges. Use expected_content_revision when paging; inspect truncation and next_view_range before requesting more.
creation_project_graph
ChatGPTRead a bounded Creation graph outline across all node types. Follow page.continuation with the same root/depth; expand frontier_node_ids with root_node_id. Use project_nodes for details and project_search to locate nodes.
creation_project_info
ChatGPTRead a Creation project and its permission context.
creation_project_list
ChatGPTList projects across systems, or select scope v2/v3. Use api_version to select the existing project tool family; never migrate implicitly. Follow next_cursor with the same scope. Do not mix page and cursor.
creation_project_migrate
ChatGPTMigrate a visible legacy project before reading content or editing. Default reuse returns an existing successful target or starts the first import. Use reimport only when the user explicitly requests a new import after understanding that all earlier projects are preserved. Retry requires the failed job ID. Reuse client_request_id for the same submission; use the returned V3 project_id for subsequent tools.
creation_project_migration_status
ChatGPTRead migration progress. On success use the returned V3 project_id; on failure explicitly retry with project_migrate mode=retry and this migration_job_id.
creation_project_nodes
ChatGPTInspect exact graph nodes, memberships, default resource, candidate IDs, drafts and runtime. At most 10 nodes; use asset_resources for more candidates.
creation_project_restore
ChatGPTRestore only after the user requested it, using the matching preview_token.
creation_project_restore_preview
ChatGPTPreview the effect of restoring a selected message and obtain a preview_token. This does not restore the project.
creation_project_runtime
ChatGPTResolve an existing project's API version before selecting its tool family. Never migrates.
creation_project_search
ChatGPTFind graph nodes by text, optionally within a container or by node type. Follow page.continuation with the same search parameters.
creation_quick_cancel
ChatGPTCancel a quick generation batch when possible; completed outputs remain available. Cancellation cannot be undone. May initialize/import history first. Charges follow actual settlement, not cancellation requests.
creation_quick_generate
ChatGPTGenerate standalone images, video clips, speech/audio or music, or revise prior quick outputs. Use when the user wants media assets directly; no project creation or Agent run is needed. For project-scoped creation or revisions, storyboards, multi-shot planning or timeline assembly, use creation_run. Choose by the delivery goal and project context, not media type or duration. Pass uploaded media as inputs[].file_ref; use creation_generation_models for model parameters. May initialize/import quick history first; follow next_step then retry with the SAME client_request_id. Use a NEW request ID only for a new generation. Historical sources inherit omitted recipe fields.
creation_quick_history
ChatGPTRead paginated quick generation history. First access may initialize/import old history. Reuse task/resource IDs for generation; keep filters unchanged when paging.
creation_quick_migration_retry
ChatGPTExplicitly retry failed/cancelled quick history preparation. Reuse client_request_id if the response is lost; never automatically loop on migration failures.
creation_quick_migration_status
ChatGPTRead preparation status; after success retry the ORIGINAL operation. This call never starts a migration or generation.
creation_quick_status
ChatGPTRead quick generation tasks and signed media results by batch, task IDs, or original request ID. May initialize/import quick history first. Follow next_step while processing. In Codex, open creation_media_preview once the first item is ready; include the whole batch.
creation_resource_info
ChatGPTRead one resource and its permitted media URL.
creation_resource_status
ChatGPTRefresh up to 100 media tasks in this project independently of run event cursors. Use task IDs from run progress; these are explicitly selected project tasks.
creation_run
ChatGPTDelegate a project-scoped creative request or revision to the Creation Agent: planning, storyboards, coordinated shots and audio, timeline assembly, or changes to an existing project's content. For standalone images, video clips, speech/audio or music, including revisions of quick outputs, use creation_quick_generate. Keep revisions to an existing project in that project. If the delivery goal is unclear, clarify whether the user wants a standalone asset or a composed video. Pass uploaded file_refs in files. Reuse client_request_id when retrying the same submission; track the returned conversation and stream IDs. To replace a pending question with a new request, pass its supersede_pause_id. Approval cannot be bypassed.
creation_run_cancel
ChatGPTRequest cancellation of the exact stream and its pending work. Cancellation cannot be undone. It may still be in progress; poll run_status until terminal. Completed work is preserved and consumed credits are not refunded by cancellation.
creation_run_result
ChatGPTRead all available events and resource states for the selected run, plus the latest 100 conversation messages and pending actions. No event pagination is required. In Codex, use media_preview to display successful resources.
creation_run_resume
ChatGPTAnswer a server-issued pending question using its action_id; ordinary feedback goes to creation_run.
creation_run_status
ChatGPTRead all available progress since from_message_id without pagination. Pass next_event_id as from_message_id on the next poll; keep the same conversation_id and stream_chat_id. Refresh previously observed media tasks using resource_status even after their events have been consumed. In Codex, open media_preview once the first resource succeeds, passing this run identity.
creation_sequence_move
ChatGPTMove one existing Sequence member after another, or to the first position with after_member_id=null. Pass expected_container_revision from project_nodes; on conflict reread before retrying. Changes Sequence order only, not Composition clip timing.
creation_skill_categories
ChatGPTList Skill categories.
creation_skill_document_get
ChatGPTRead an owned Skill's editable Markdown and revision.
creation_skill_feed
ChatGPTBrowse community Skills.
creation_skill_get
ChatGPTRead a Skill, owner draft, or privately shared Skill.
creation_skill_install
ChatGPTInstall an accessible Skill. May disable another Skill when the enabled-Skill quota is full.
creation_skill_list
ChatGPTList the caller's Skills.
creation_skill_reference_get
ChatGPTRead an owned draft reference by node_id with content revision. May create its private editing workspace on first access.
creation_skill_reference_list
ChatGPTList reference summaries of an owned Skill draft. May create its private editing workspace on first access. Read skill_document_get for the current draft revision.
creation_skill_reference_read_live
ChatGPTRead a live reference by name from an accessible Skill. Use shared_token for private sharing and expected_version to pin the live snapshot.
creation_skill_uninstall
ChatGPTDisable a Skill for the caller without deleting its content or removing library membership. This is reversible: use creation_skill_install to enable it again.
creation_upload_complete
ChatGPTFinish an upload_id and obtain file_ref. Pass media references to creation_quick_generate.inputs[].file_ref for standalone generation, or materials to creation_run.files for project work. Files may be validated, moderated or converted; document OCR or model-based parsing may continue in the background after file_ref is available for referencing. Retry with the same upload_id; processing may repeat until completion is recorded. file_ref_expires_at is a Unix timestamp.
creation_upload_ticket
ChatGPTPrepare a direct upload; retries may create a new upload_id. Files may be validated, moderated or converted; documents may require OCR or model-based parsing, including background processing, even for a deduplicated upload. For video, first probe the actual file (for example with ffprobe) and provide video_metadata: width and height in pixels, duration in seconds, and has_audio (false if no audio stream). Do not guess metadata. Use transport, then call creation_upload_complete.