append_generated_image_upload_chunk
ChatGPTAppends one sequential base64 chunk to a workspace-scoped generated-image upload. The uploadId, part number, and byte offset identify the next position; chunk size is capped to fit the MCP request limit.
approve_post_variant
ChatGPTApproves one exact saved Post channel version after the user explicitly asks to approve it. Inputs bind postId, expectedPostRevision, variantId, expectedVariantRevision and expectedContentHash to the current list_posts/get_context result; confirmation=CONFIRM_APPROVE_POST_VARIANT reflects that user instruction. The server rejects stale IDs, revisions or content hashes, verifies publication readiness and excludes an in-progress delivery under the normal review lock. This changes editorial approval only; it does not schedule or publish. Requires the separate review permission. A unique idempotencyKey identifies this requested decision and is reused for exact retries.
archive_all_drafts
ChatGPTStages every current unscheduled draft shown in Publish > Drafts for reversible archive confirmation. The server selects targets from workspace metadata and returns the target count with a review receipt; draft text, titles, target IDs, and source provenance are not exposed. Google-derived and other restricted posts are included by this metadata-only operation. A private in-app chat shows one Yes/Cancel confirmation card; no change occurs until confirmation. Posts with schedules or active planned slots are excluded, and Apply checks draft and placement state again. If more than 100 drafts match, nothing is staged and the result reports the limit. Archive affects local SuperPost objects only, not social-network posts. For an external MCP call, supply a fresh idempotencyKey for each new user-intended archive request, including after Cancel or Reject; reuse the same key only when retrying that request after an uncertain response. Omitting it preserves legacy deduplication and can replay an earlier result.
archive_idea
ChatGPTStages a verified idea for archive confirmation in the current SuperManager chat. In-app requests group staged ideas into one confirmation card; the idea changes only after the user confirms. Archive is reversible and does not permanently delete the idea. External MCP returns the existing proposal review flow.
archive_post
ChatGPTStages an exact, workspace-verified SuperPost Post or draft for reversible archive confirmation in the current SuperManager chat. In a private in-app chat, no change occurs until the user confirms; Cancel leaves it unchanged. Archive affects only the local SuperPost object and does not delete a social-network post. External MCP returns the existing SuperManager review proposal.
archive_selected_posts
ChatGPTStages the exact supplied SuperPost Post IDs and revisions for reversible archive confirmation. All targets are rechecked in the workspace before staging. The operation scope is exactly the supplied selection and is not expanded based on status. A private in-app chat shows one confirmation card; no change occurs until confirmation. External MCP creates the existing review proposal. Archive affects local SuperPost objects only, never deletes social-network posts, and publication/slot safety checks still apply.
begin_generated_image_upload
ChatGPTBegins a workspace-scoped upload session for generated image bytes. Supported formats are PNG, JPEG, and WebP up to 10 MiB. The session accepts sequential byte chunks and finalization produces a temporary Media ID for a Post media request. URL and video uploads are unsupported.
create_content_bundle
ChatGPTDirectly saves a new, exact content bundle as unscheduled Post drafts in SuperPost after validating the same revision, source, media, workspace and business rules used by the review flow. It can create new Ideas and Posts in the exact bundle, but cannot edit existing Posts or Ideas. The returned receipt contains IDs and statuses only, not post text. Requires the separate draft permission and confirmation=CONFIRM_SAVE_DRAFTS as a reflection of the user’s explicit request; do not ask the user to confirm again.
create_idea
ChatGPTCreates a new idea in the workspace. An AI-authored concept uses origin=ai_generated and includes generationNotes, a concise user-facing explanation of the actual reason and angle for the idea. This note contains no private chain of thought, unsupported claims, or invented evidence. A user-supplied concept uses origin=user; generationNotes is not accepted for user-authored ideas. SuperPost stores the submitted note in the idea's Why this idea receipt without inventing model details or evidence. idempotencyKey identifies one intended creation; retries with the same key deduplicate, while a distinct key represents a separate creation even when its content is identical. Without a key, a retry may create a duplicate. Client-supplied server-verified sourceProvenance is not accepted; external citations are labeled imported/unverified.
decide_immediate_publications
ChatGPTApplies approve or reject to supplied existing pending_approval publication jobs whose dispatchMode is immediate. Exact scheduleId, expectedUpdatedAt and expectedPublicationHash values correspond to list_pending_publications; get_context returns full Post text when a preview is truncated. The confirmation marker is specific to the selected decision and the request carries the user's choice for this exact batch. Approval applies SuperPost's existing workspace policy and releases the same queued job for immediate dispatch; rejection cancels that job. No new or replacement job is created. Revisions, content hashes, workspace ownership and current policy are rechecked, and results report each item independently. Each existing schedule is atomically resolved from pending_approval, so an exact retry reports already_resolved instead of resolving it again; idempotencyKey is a request correlation label, not a persisted decision ledger. Distinct batches have fresh keys and an exact retry reuses the same key. Requires the separate publish permission.
decide_scheduled_publications
ChatGPTApplies approve or reject to supplied existing pending_approval publication jobs whose dispatchMode is scheduled. Exact scheduleId, expectedUpdatedAt and expectedPublicationHash values correspond to list_pending_publications; get_context returns full Post text when a preview is truncated. The confirmation marker is specific to the selected decision and the request carries the user's choice for this exact batch. Approval applies SuperPost's existing workspace policy and releases the same future job for its saved time; rejection cancels that job. No new or replacement job is created. Revisions, content hashes, workspace ownership and current policy are rechecked, and results report each item independently. Each existing schedule is atomically resolved from pending_approval, so an exact retry reports already_resolved instead of resolving it again; idempotencyKey is a request correlation label, not a persisted decision ledger. Distinct batches have fresh keys and an exact retry reuses the same key. Requires the separate schedule permission.
finalize_generated_image_upload
ChatGPTFinalizes a complete workspace-scoped generated-image upload and returns a temporary Media ID for use in a Post media request proposal. This uploads the image only; it does not create a Post or start media generation.
get_context
ChatGPTReads account-scoped context for a content variant, idea, post, assistant thread, publication, schedule, or exact source capture. A source_capture returns its exact saved text and verified provenance, including manual captures outside the bounded initial references; Google-derived or unverified source content is withheld. A post returns full shared text and platform variants only when its source provenance is verified as safe; Google-derived or unverified content is withheld. Thread and publication context is metadata-only. Idea context includes the saved assistantBrief when present and the full generationReceipt when available. assistantBrief is saved writing guidance for a future post and is separate from the thesis, conclusion, and post body. The receipt may contain model name, platform/cadence selection, concise explanation, explanation-cited evidence IDs, context evidence, and context status/hash; fields absent from the receipt were not saved and must not be inferred. MCP-authored receipts preserve the submitted user-facing generation note and omit model, platform selection, or evidence when the tool did not provide them. Cited evidence IDs are distinct from the broader providedEvidence list. The explanation is a saved rationale, not private chain of thought; older ideas may have no receipt.
get_idea_performance
ChatGPTCompares topics, labels, and linked ideas across available managed and imported history for the selected time window and scope. Optional topic/label and idea filters narrow the comparison. Provider, account, metric, and publication-age cohort remain distinct in the results. Fewer than three comparable peers are marked insufficient. sourceTruncated indicates repository read limits. Raw metrics are not compared across platforms.
get_idea_potential
ChatGPTRanks active workspace ideas by historical potential, with optional provider, level, and used-status filters plus bounded pagination. Scores are 0–100 ranking signals, not success probabilities. Topic/label results and similar successful ideas are measured within one platform, account, metric, and age cohort. The result includes platform evidence, confidence, semanticStatus, and sourceTruncated. An idea's own historical wins are excluded from its score. Semantic evidence can warm asynchronously; semanticStatus reports warming. This is read-only.
get_post_media_request
ChatGPTReads a workspace-verified Post's current image/video request, safe lifecycle state, and media capabilities. Video capabilities distinguish supported operations and output formats from runtime readiness, which is not probed by this read. Connected account targets are represented by opaque workspace-scoped targetKeys. This does not start or change media work.
get_profile
ChatGPTReturns the profile represented by this request's authenticated SuperPost credentials. The opaque ID stays stable across token refresh and reconnection.
get_source_image
ChatGPTReads the status of a source image job bound to sourceAssetId and expectedRevision. A completed result includes the original workspace mediaId for reuse in an authorized media request and a bounded JPEG view as native image content. The view is a reduced derivative, not the full-resolution original. Current source grants, provider, revision, job binding, and media ownership are rechecked. This read does not initiate or retry work, publish content, or modify source files.
get_source_transcription
ChatGPTReads the status of a transcription job bound to an exact sourceAssetId and expectedRevision. The current source grant and revision are checked again. A completed current-revision transcript is returned with a 12,000-character limit; old-revision transcript text is withheld. Status reads do not start or retry work.
import_ideas
ChatGPTBulk-import ideas with multi-source provenance. Each source is labeled imported/unverified. Requires import scope. Deterministic IDs enable idempotent replay.
import_linkedin_analytics
ChatGPTValidate or import a LinkedIn Creator Analytics XLSX export for a connected account. Send the XLSX bytes as canonical base64 in fileBase64. Set dryRun=true to preview without saving; imports replace an earlier export for the same account and date range. This imports period totals and daily aggregates; it does not create lifetime post analytics.
link_linkedin_analytics_post
ChatGPTAssociates one exact post URL from an imported LinkedIn Creator Analytics XLSX report with existing workspace ideas. The URL must appear in the selected channel's saved report, and every idea must belong to this workspace. The operation adds report-to-idea analytics links only; it does not alter idea provenance or publication records.
list_ideas
ChatGPTReads the workspace idea catalog with optional generic text search, source provenance search, topic, label, and cursor-based pagination. The sourceQuery filter matches source author, handle, source title or filename, provider, remote item ID, or URL; English/Russian transliterations match, and | separates alternate spellings. Provenance search does not match names mentioned only in idea text or source excerpts. Returned source attributions are bounded. This is read-only.
list_linkedin_analytics_channels
ChatGPTLists connected LinkedIn channels available for an analytics XLSX import. Each result includes a channelId identifying the account, including when multiple LinkedIn channels are connected.
list_pending_publications
ChatGPTLists existing publication jobs waiting for manual approval in this workspace. It returns each exact schedule ID, updatedAt revision and payload hash, the dispatch mode/time, connected-channel label, and the bound Post/variant IDs and content hash. Short content fields are included only when the existing MCP post-content disclosure allows them; full Post text can be checked with get_context using the returned postId. The binding, contentStatus and preview-truncation fields describe whether the visible snapshot is complete and current. Pagination exposes a nextCursor for subsequent reads whenever hasMore is true. This is a read of the existing queue and never creates or approves work.
list_posts
ChatGPTSearches saved posts with bounded pagination. Results contain compact canonical references only; Google-derived or unverified post text is withheld before matching. Full shared post text and platform versions are not included in search results.
list_publications
ChatGPTList recent published receipt metadata, optionally filtered by a verified post ID or provider. Publication content excerpts are never returned.
list_source_scopes
ChatGPTLists the signed workspace's connected source folders and file scopes. Returns exact scope IDs, display paths, include/exclude patterns, and updatedAt revisions; provider connection IDs and credentials are omitted. Scope IDs are workspace-scoped selectors. Paginated results include nextOffset.
list_threads
ChatGPTList assistant conversations using bounded cursor pagination. Returns neutral titles and latest run status only; it does not expose conversation messages or source-derived titles.
preview_content_bundle
ChatGPTValidates a complete bundle of new Ideas, post drafts, and supported links without saving the content objects. Each Post has a unique key, sharedText, one or more platform variants, and supported source/media relationships; variants belong to the same Post. A mediaRequest is separate from Post text. mediaRequest.inputSources contain sourceAssetId, sourceRevision, and a required_material or reference role. mediaRequest.targetChannelKeys contain opaque workspace-scoped channel selectors. Existing IDs are validated and source revision guards are recorded. Generated-image attachments use workspace Media IDs; arbitrary media URLs and paths are rejected.
preview_content_variant_edit
ChatGPTPreview a bounded content variant edit without applying it.
preview_idea_edit
ChatGPTPreviews a bounded idea edit without applying it. Supported edits include replace_idea_assistant_brief, which stores writing guidance for a future post separately from the thesis, conclusion, and post body.
preview_post_media_request_change
ChatGPTPreviews a revision-guarded Post media request change. Connected social-channel source assets are selected with exact workspace-scoped targetKey values in targetChannelKeys; raw connection IDs are rejected. The preview is read-only; saving or starting image/video work requires application of a later proposal.
preview_profile_instruction_change
ChatGPTPreviews an exact account or connected-channel editorial-guidance change. Account guidance supports up to 2,000 characters; channel guidance up to 500. A channel edit requires an authorized provider and workspace-scoped targetKey. The result includes before/after text and a version guard; no change is saved.
preview_profile_settings_change
ChatGPTValidates a typed profile, Autopilot, or structured writing-standards proposal against the latest saved settings. Structured antiSlop/contentStructure values are available in profile instructions. Writing scopes are writing_settings_account or writing_settings_channel. Proposals contain only changed fields; omitted fields retain their saved values. Null selects the product default for account settings or inheritance for channel settings; an empty string explicitly disables a field. Channel proposals require the returned targetKey and expectedProfileVersion as well as expectedVersion. The request includes the exact scope, selector, before values, and version guards. The result contains before/after values. Validation does not save settings, start Autopilot, or publish content. Secret entry, provider login, and consent disclosures use secure handoffs.
preview_publish_posts_now
ChatGPTValidates an exact batch of saved Posts and returns a snapshot of each destination, exact copy, media identity, revision and content-approval status. It does not approve, enqueue, schedule, or publish anything. The previewHash binds that exact content and destination. If content is not yet approved, the publish action will approve only this exact previewed content. Requires the separate publish permission.
preview_schedule_edit
ChatGPTPreview a bounded publication schedule edit without applying it.
preview_schedule_posts
ChatGPTValidates an exact batch of saved Posts and returns a snapshot of each destination, copy, media identity, revision and future time, plus a previewHash. This does not approve, schedule, or publish anything. The previewHash binds the current saved content and timing, and the separate apply action requires that exact hash after the user explicitly authorizes the exact future schedule. Requires the separate schedule permission.
preview_source_scope_rules
ChatGPTPreviews replacement include/exclude rules for one source scope. The request identifies the scope by sourceScopeId and its current updatedAt revision as expectedUpdatedAt. Each supplied array replaces that full list; an omitted field remains unchanged and [] clears the list. Patterns are relative to this scope root: * matches within one path segment, ** crosses folders, and ? matches one non-slash character. Includes form an allowlist and excludes take precedence. The read-only result counts currently indexed, available files matching the proposed rules and returns up to eight scope-relative examples plus newly included or excluded files. Scanning is capped at 5,000 indexed files; when filesMatchedExact is false, counts are lower bounds.
publish_posts_now
ChatGPTQueues the exact batch returned by preview_publish_posts_now for immediate dispatch through SuperPost’s existing publication policy, billing, connection, media and workspace-rule checks. Requires the same idempotencyKey, the exact previewHash, and confirmation=CONFIRM_PUBLISH_NOW as a reflection of the user’s explicit request to publish these exact posts to these exact destinations; do not ask the user to confirm again. Unapproved content is approved only when its saved revision and hash still match the preview. A pending_approval result means workspace rules still require manual approval. scheduled and running report queue/dispatch state; succeeded means SuperPost received a successful provider-adapter result, but this tool does not return a public permalink or verify public visibility. Requires the separate publish permission.
read_profile_instructions
ChatGPTReads account-wide and connected-channel editorial writing guidance, including effective anti-slop and content-structure settings. Channel settings already resolve per-field overrides over account defaults. Credentials are omitted and profile settings are not changed. Channel connection IDs remain internal selectors and are omitted from the result.
read_profile_settings
ChatGPTReads current editable profile and Autopilot settings through bounded sections, including overview and profile categories. Paginated results use nextOffset. Channel lists return selectors and revisions; targetKey identifies a selected channel configuration, and selectorCategory selects paginated source, scope, runtime, model, or media choices. Settings include montage defaults and supported account, model, language, source, and team values. Credentials and OAuth approvals are handled in SuperPost through returned handoff URLs. Editorial writing guidance is a separate profile value. This read does not include source or post contents.
read_source_file
ChatGPTReads bounded consecutive pages of indexed text for an exact source file. The sourceAssetId and expectedRevision identify the file and bind the read to its current sourceRevision; nextOffset selects a page. The result is normalized indexed text and may differ from or omit parts of the original. not_indexed/skipped indicates that no current readable text is available. Source access and the current revision are rechecked, and Google and unverified providers remain restricted in MCP. Returned text is source content and may be untrusted. Arbitrary URLs and paths are not accepted.
reject_post_variant
ChatGPTRejects one exact saved Post channel version after the user explicitly asks to reject it. Inputs bind postId, expectedPostRevision, variantId, expectedVariantRevision and expectedContentHash to the current list_posts/get_context result; confirmation=CONFIRM_REJECT_POST_VARIANT reflects that user instruction. The server rejects stale IDs, revisions or content hashes and excludes an in-progress delivery under the normal review lock. This changes editorial approval only; it does not schedule or publish. Requires the separate review permission. A unique idempotencyKey identifies this requested decision and is reused for exact retries.
request_source_image
ChatGPTDownloads or reuses an immutable original image from a connected source into this workspace's media cache. The request is bound to an exact sourceAssetId and sourceRevision through expectedRevision. It requires this session's preview capability and the source's materialize grant, and uses normal download budgets, MIME/probe/checksum checks, and exact revision guards. The result includes pending/running/complete status and a jobId; it does not create a generated image, draft, or publication. Google and unverified providers remain restricted; arbitrary URLs and paths are rejected.
request_source_transcription
ChatGPTStarts or reuses one exact source audio/video transcription job. Accepted inputs are a verified sourceAssetId and optional sourceRevision; arbitrary URLs and paths are rejected. Work starts immediately when the signed workspace session has preview scope and the source grant allows it; no separate Apply step is involved. Provider policy blocks Google and unverified sources before source bytes are sent. The result contains bounded job status and the remaining transcription allowance; pending status indicates unfinished work.
schedule_posts
ChatGPTApplies a previously returned future-schedule preview by creating scheduled slots in SuperPost. Requires the exact previewHash, the same idempotencyKey as that preview, and confirmation=CONFIRM_FUTURE_SCHEDULING. This action requires the user’s explicit authorization for these exact posts, destinations, and future times; OAuth permission alone does not authorize a particular schedule. For an unapproved post, approveExactContent is required for that exact item only when the user explicitly authorized its content. Scheduling does not publish immediately; SuperPost publishes at each selected future time. The server rechecks all content revisions, hashes, provider targets, media, channel mapping, and time before applying the batch.
search_sources
ChatGPTSearches relevant excerpts from the account's permitted sources. This read-only operation does not change or publish data.
submit_content_bundle_proposal
ChatGPTSubmits an exact content bundle to SuperPost as a review proposal. It does not create the proposed ideas or posts, schedule them, or publish them. A mediaRequest is separate from Post text. inputSources contain verified sourceAssetId, sourceRevision, and role values; targetChannelKeys contain exact workspace-scoped channel selectors. Generated-image attachments use workspace Media IDs; arbitrary URLs and paths are rejected.
submit_content_variant_proposal
ChatGPTSave a validated content variant edit in SuperPost for the user to review and explicitly apply or reject.
submit_idea_proposal
ChatGPTSave a validated idea edit in SuperPost for the user to review and explicitly apply or reject.
submit_post_media_request_proposal
ChatGPTSubmits a validated image/video request change for review. In private in-app chat, no configuration changes or paid media jobs start until the user applies the review card; external MCP uses its existing review flow. Generated-image requests accept workspace Media IDs in inputMediaIds. Arbitrary media URLs and raw connection IDs are rejected; targetChannelKeys accepts exact workspace-scoped channel selectors.
submit_profile_instruction_proposal
ChatGPTSubmits an exact before/after editorial-guidance change with expectedVersion as a review proposal. Account guidance supports up to 2,000 characters; channel guidance up to 500. Private in-app changes appear in the current chat for Apply/Reject; external MCP uses the existing review flow. The change is not saved until the user applies it.
submit_profile_settings_proposal
ChatGPTSubmits a typed profile, Autopilot, or structured writing-standards proposal for Apply/Reject review; the setting is not saved until application. Writing-standard proposals preserve omitted fields and distinguish null (product default or channel inheritance) from empty string (explicitly disabled). Current user/workspace permissions and saved versions are checked again on Apply. Autonomy, publishing permissions, budgets, source choices, disconnects, and team actions are guarded by explicit review. Credential values are not accepted; secure handoffs handle secret entry, provider login, and consent disclosures.
submit_schedule_posts_proposal
ChatGPTCreates one reviewable batch for scheduling saved Post variants across connected social channels. Each item identifies an exact postId and variantId, an opaque targetKey from the channel list, a provider-compatible publication target, localDateTime, and IANA timezone. Publish text and media are captured with the Post revision, variant revision, and content hash in the SuperPost review proposal; the external MCP response contains only a review receipt and count. Already-approved Post and variant content is eligible by default. The approveExactContent option represents explicit user authorization for the exact copy shown in the SuperPost review card, where approval and future scheduling take effect together on Apply. Proposal submission alone does not approve or schedule anything. Direct scheduling requires a separate permission and the user’s explicit authorization for exact future times. The idempotencyKey identifies the logical request; retries return the same review proposal, while different requests use a different key. Duplicate variants, duplicate Post/destination/time slots, and variants with an active schedule are rejected. Email campaign targets (Brevo, Mailchimp, Kit, and Listmonk email-send) are outside this Post flow because they schedule provider-side campaigns rather than saved Post variants.
submit_schedule_proposal
ChatGPTSave a validated schedule edit in SuperPost for the user to review and explicitly apply or reject.
update_source_scope_rules
ChatGPTPersists file/folder wildcard rules immediately for one selected source scope explicitly requested by the user; no separate confirmation card is created. The request requires sourceScopeId, expectedUpdatedAt, complete desired arrays, and previewHash. A revision or preview mismatch rejects the save. The target is restricted to the authenticated workspace and exact selected source scope; connection IDs and cross-workspace selectors are not accepted. A successful update immediately governs source selection, and the next sync reconciles indexed coverage. Patterns are relative to the scope root: * matches one path segment, ** crosses folders, and ? matches one non-slash character. Excludes take precedence over includes.
approve_post_variant
Claudearchive_all_drafts
Claudearchive_idea
Claudearchive_post
Claudearchive_selected_posts
Claudecreate_content_bundle
Claudecreate_idea
Claudedecide_immediate_publications
Claudedecide_scheduled_publications
Claudeget_context
Claudeget_idea_performance
Claudeget_idea_potential
Claudeget_post_media_request
Claudeget_source_image
Claudeget_source_transcription
Claudeimport_ideas
Claudeimport_linkedin_analytics
Claudelink_linkedin_analytics_post
Claudelist_ideas
Claudelist_linkedin_analytics_channels
Claudelist_pending_publications
Claudelist_posts
Claudelist_publications
Claudelist_source_scopes
Claudelist_threads
Claudepreview_content_bundle
Claudepreview_content_variant_edit
Claudepreview_idea_edit
Claudepreview_profile_instruction_change
Claudepreview_profile_settings_change
Claudepreview_publish_posts_now
Claudepreview_schedule_edit
Claudepreview_schedule_posts
Claudepreview_source_scope_rules
Claudepublish_posts_now
Clauderead_profile_instructions
Clauderead_profile_settings
Clauderead_source_file
Claudereject_post_variant
Clauderequest_source_image
Clauderequest_source_transcription
Claudeschedule_posts
Claudesearch_sources
Claudesubmit_content_bundle_proposal
Claudesubmit_content_variant_proposal
Claudesubmit_idea_proposal
Claudesubmit_profile_instruction_proposal
Claudesubmit_profile_settings_proposal
Claudesubmit_schedule_posts_proposal
Claudesubmit_schedule_proposal
Claudeupdate_source_scope_rules
Claude