append_brand_file_upload_chunk
ChatGPTPro owner only. Append one exact indexed chunk of at most 256 KiB to the active brand-file upload session, with its current upload revision and the chunk's lowercase SHA-256. Send only the selected file's base64 bytes. Does not complete or register the file. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
begin_brand_file_upload
ChatGPTPro owner only. Begin one consented brand-file upload session, up to 32 MiB, using the exact current snapshot hash, complete size and lowercase SHA-256. The owner/brand-bound session expires after 30 minutes. Does not upload chunks or register an asset. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
cancel_brand_file_upload
ChatGPTPro owner only. Cancel one explicitly selected incomplete upload session and discard its staged chunks, using the current upload revision. Does not delete an already registered brand file. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
complete_brand_file_upload
ChatGPTPro owner only. Complete one uploaded brand file at its current session revision and current Brand Page snapshot hash. Verifies full size and SHA-256 before registering the stored file. Does not add a Brand Page item or publish a release. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
create_brand_custom_section
ChatGPTPro owner only. Create one custom Brand Page section with its authored content and current section hash. Does not modify built-in section content or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
create_brand_illustration_reference
ChatGPTPro owner only. After the user's explicit request, register one uploaded illustration file as style or imagery inspiration in the active Brand Page draft. First call upload_brand_file with file_kind illustration; then pass its file_id here. The server sets a reference-only type, so agents must not use the image as a direct brand asset. Read list_brand_page_items with resource illustrations to verify. This does not publish a Brand Release.
create_brand_interface_component
ChatGPTPro owner only. Create one authored Interface System component as a draft. Pass the current Interface System hash and complete component definition. Does not manufacture verified source or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
create_brand_interface_variant
ChatGPTPro owner only. Create one authored variant on an exact Interface System component using its current hash. Does not invent a captured Figma state or verified implementation. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
create_brand_page_item
ChatGPTPro owner only. After the user's explicit request, create one color, typography, logo, font or illustration in the active Brand Page. File-backed items require a file_id returned by upload_brand_file. Colors accept name and hex; typography requires style_type (title, body or caption); logos require slot and file_id; fonts and illustrations require file_id. For a visual reference, prefer create_brand_illustration_reference so the asset cannot be mistaken for an illustration to use directly.
create_brand_snapshot_item
ChatGPTPro owner only. Create one snapshot item in the selected resource. Read list_brand_snapshot_items first; pass its exact snapshot hash and a new item. Does not edit an existing item or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
delete_brand_custom_section
ChatGPTPro owner only. Delete one explicitly selected custom Brand Page section and its content, using its exact ID and current section hash. Built-in sections must be hidden instead. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
delete_brand_design_guidance
ChatGPTPro owner only. Delete one explicitly selected spacing, rounding or textual-rule row by exact ID and current guidance hash. Built-in presets are reset according to their existing lifecycle. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
delete_brand_file
ChatGPTPro owner only. Physically delete one explicitly selected, unreferenced file belonging to the active brand. Requires the current snapshot hash and confirm_physical_delete:true. Any editor, snapshot, Interface System or immutable release reference blocks deletion. Removing a Brand Page item alone never authorizes physical file deletion.
delete_brand_interface_component
ChatGPTPro owner only. Remove one explicitly selected component from the active Interface System using its exact ID and current hash. Preserves imported source according to the existing component lifecycle. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
delete_brand_interface_variant
ChatGPTPro owner only. Delete one explicitly selected authored component variant by exact component ID, variant and current Interface System hash. Existing source-preservation restrictions apply. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
delete_brand_page_item
ChatGPTPro owner only. After an explicit user deletion request, delete one exact editor item and its exact snapshot counterpart. Read the items and draft first; pass item_id and current snapshot_content_hash. Uploaded files remain stored. Ambiguous matches or stale hashes block deletion. Reload both stores after a reconciliation error before retrying.
delete_brand_snapshot_item
ChatGPTPro owner only. Delete one explicitly selected snapshot item using its exact ID and current snapshot hash. Editor-backed assets use their dedicated lifecycle. Does not delete stored files or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
fetch
ChatGPTOptional document helper. Fetch a known read-only brand document by ID. Do not call this before get_brand_context unless a document ID is already known.
get_brand_assets
ChatGPTCall directly to get complete official URL-only logos, fonts, colors, illustrations, patterns, iconography, imagery and additional assets, including authored metrics, rules, constraints and delivery metadata. Search or fetch is not required first.
get_brand_context
ChatGPTCall directly before brand-aware work. Returns every active Brand Page object with its authored fields, URL-only asset delivery, section lifecycle status and the normalized Brand Contract V2 object inventory with exact source paths, rules, enforcement, coverage and missing fields. Archived sections are preserved but unavailable to agents; search or fetch is not required first.
get_brand_editor_capabilities
ChatGPTPro owner only. Read the current server's complete Brand Page authoring capability matrix, file limits and source-preservation boundaries. This does not edit or publish the brand. Use it to detect stale host tool catalogs before editing.
get_brand_file_upload_status
ChatGPTPro owner only. Read the status and current revision of one upload session belonging to this owner and brand. Does not upload chunks, complete or cancel the session. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
get_brand_interface_draft
ChatGPTPro owner only. Read the authoritative Interface System component and variant authoring draft, governance, source eligibility and content_hash. Imported source and implementation evidence remain distinct from authored metadata. This does not approve, compile or publish components.
get_brand_page_draft
ChatGPTPro owner only. Read the authoritative Brand Page draft and snapshot_content_hash before editing. Includes profile, color behavior, layout contract, grids, do/don't rules, colors, logos, typography, iconography, voice and output rules.
get_brand_reference_for_output
ChatGPTReturns the output brand package: all approved colors, typography, layout, rules, complete reference-only visual set and the resolved logo. Supply actual logo background and space. For a frontend output, this normally replaces a separate get_output_reference call. Reference-only images guide new original imagery and cannot be placed directly. Does not generate or render files.
get_interface_component_implementation
ChatGPTCall for every selected approved Figma Interface System component before editing UI. Returns the immutable official React artifact only when every imported state has self-contained SVG with text converted to vector outlines. The artifact records its source hash and variant count, and includes byte-for-byte files with SHA-256 hashes. Use captured variants/states and customize only the outer wrapper; visible text and internal paths stay as imported. Write every file at its declared path without modifying it; a path, export name, component ID or hand-built substitute is never sufficient evidence.
get_output_reference
ChatGPTCall after a ready preflight for output-specific brand reference instructions. Returns official asset catalogs, visual_execution, typography metrics, color behavior, layout, visual references, voice, brand feel and saved rules; it never renders, composes or exports files. Inspect reference-only images as visual inputs before generating new imagery. For a specific logo, call get_brand_reference_for_output with actual background and space.
get_perture_settings
ChatGPTPro owner only. Read agent behavior and Figma plugin sync settings for this account.
get_work_memory
ChatGPTCall directly to return compact active Work Memory for the connected brand. It never returns raw feedback history or full output artifacts.
hide_brand_page_section
ChatGPTPro owner only. Hide one explicitly selected Brand Page section while preserving its content. Use the current hash from list_brand_page_sections. Hidden content is unavailable as active agent guidance. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
import_brand_interface_source
ChatGPTPro owner only. Import the user's explicitly selected component source through Perture's existing Figma import validation. Requires the current Interface System hash and complete consented source payload. Source may replace matching components; host-supplied data is not a live Figma-capture attestation. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
link_legacy_brand_logo
ChatGPTPro owner only. Link one exact legacy snapshot logo to the same durable source in an existing editor logo, using the current snapshot hash. Replaces the selected association; source identity must match. Does not delete a file or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
list_brand_design_guidance
ChatGPTPro owner only. Read the active Brand Page's Interface System spacing and rounding measurements and content_hash before editing.
list_brand_files
ChatGPTPro owner only. List bounded metadata for files belonging to the active brand, including file IDs needed for replacement or explicit cleanup. This does not expose file bytes or declare a file unreferenced. Physical deletion performs a separate authoritative reference and publication check.
list_brand_page_items
ChatGPTPro owner only. Read editable Brand Page items and their IDs before making a scoped change. For illustrations, each item also reports computed usage_mode, reference_only and use_official_asset_directly; these describe the draft, not the published release.
list_brand_page_sections
ChatGPTPro owner only. Read Brand Page section IDs, visibility, ordering, custom content and current Interface System content_hash. Hidden section content is preserved rather than applied as active brand guidance.
list_brand_snapshot_items
ChatGPTPro owner only. Read authoritative snapshot items and snapshot_content_hash for grids, do/don't, iconography, imagery, patterns, assets, typography metrics, legacy spacing/composition or brand feel. Hidden sections stay unavailable. Read before a scoped create, edit, delete or reorder.
list_brands
ChatGPTList all Perture brands the connected account can access.
list_frontend_corrections
ChatGPTList only correction requests that the signed-in user explicitly created in Perture Logs. This tool never creates a request.
plan_brand_interface
ChatGPTPlan the whole generated interface for any brand: every requested component and semantic role, plus colors, typography, layout, spacing, radii, logos, iconography and authored rules. Resolve approved executable components or mark brand-derived roles explicitly. Use before coding any frontend output.
plan_interface_control
ChatGPTLegacy single-control helper. Plans one button or CTA only; it does not cover containers, typography, grid or the whole output. Use plan_brand_interface for frontend work.
preflight_brand_output
ChatGPTCall before creating an external output. Checks the published brand, allowed colors and required capabilities. A reference-only image can authorize a new original illustration, never direct placement of that file. Follow visual_execution; if the host cannot inspect references and generate an image, stop instead of substituting a generic visual.
publish_brand_release
ChatGPTPro owner only. Publish the active Brand Page draft after the user's explicit request. First read get_brand_page_draft and pass its exact snapshot_content_hash; optionally pass the published release number for an additional conflict check. Returns the verified immutable Brand Release. A conflict requires fresh reads before retrying.
remove_legacy_brand_logo
ChatGPTPro owner only. Remove one explicitly selected obsolete legacy snapshot logo row using its exact ID and current snapshot hash. Does not delete the editor logo or stored file, and does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
reorder_brand_design_guidance
ChatGPTPro owner only. Replace the order of one guidance collection with its complete item IDs and current guidance hash. Does not change the rows' content or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
reorder_brand_interface_components
ChatGPTPro owner only. Replace the visible component order with the complete component IDs and current Interface System hash. Does not modify source or component definitions. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
reorder_brand_interface_variants
ChatGPTPro owner only. Replace one component's variant order with the complete variant names and current Interface System hash. Preserves variant definitions and imported source. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
reorder_brand_page_sections
ChatGPTPro owner only. Replace the Brand Page section order with the complete section IDs, using the current hash from list_brand_page_sections. Preserves section content and visibility. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
reorder_brand_snapshot_items
ChatGPTPro owner only. Replace the order of one snapshot resource with the complete ordered IDs. Read its current items and snapshot hash first. Does not create, edit or delete items. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
replace_brand_item_file
ChatGPTPro owner only. After an explicit replacement request, replace the file of one exact logo, font or illustration while retaining its item ID, metadata and snapshot binding. Upload the new file first, read the current draft and pass its snapshot hash. Old files remain available to immutable published releases. Verify the returned save; never retry an unknown outcome blindly.
resolve_interface_component
ChatGPTResolve a requested dotted UI intent and its background/surface to one visible, approved Brand Page component, variant and verified implementation target. Unmapped results may include discovery_candidates from catalog names or categories; these are hints only, never a selected or verified implementation. For an unmapped structural/content intent with no applicable approved component, the host may build a local composition from compatible patterns and report not_verified Interface System coverage. needs_context, no_match and ambiguous do not authorize guessing or substitution. Use only a resolved result in an official interface usage manifest entry.
restore_brand_interface_component
ChatGPTPro owner only. Restore one previously removed Interface System component by exact ID and current hash. Does not create source, declare verification or publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
save_brand_design_measurement
ChatGPTPro owner only. Save one spacing or rounding measurement in the active Brand Page after the user's explicit request. First call list_brand_design_guidance, then pass its exact content_hash. Use an existing measurement ID to edit it, or omit ID to add a custom measurement. A conflicting save must be reviewed against fresh guidance before retrying.
save_brand_interface_selection
ChatGPTPro owner only. Explicitly map any dotted interface intent to one visible approved component, exact variant, context and framework. Read list_brand_design_guidance for inventory and content_hash first. An approved Figma draft with source_authoring.status verification_on_save can be mapped even when its cached verified_targets list is empty: this tool compiles its current immutable source to verify the implementation before saving. No preliminary Brand Release is needed. Conflicting or ambiguous selections and invalid source are rejected. Preserves original source and other mappings. Does not publish a Brand Release.
save_brand_textual_rule
ChatGPTPro owner only. Create or edit one Interface System textual rule after the user's explicit request. Read list_brand_design_guidance first and pass its exact content_hash. Use an existing rule ID to edit it or omit ID to add one. Preserves all other Interface System content and verifies the save.
search
ChatGPTOptional discovery helper. Search the brand connected to this API key. Do not call this before get_brand_context unless search is specifically needed.
set_active_brand
ChatGPTRequired before calling any Perture brand tool. Call this only when the user explicitly names or asks to switch to a brand. It returns brand_lock_token; pass that token to every subsequent brand tool in this chat.
set_brand_interface_governance
ChatGPTPro owner only. Set the governance status of one exact component using the current Interface System hash. This can replace its active approval status; implementation eligibility is still validated separately. Does not publish a release. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
show_brand_page_section
ChatGPTPro owner only. Restore one hidden built-in Brand Page section so its existing items are visible. Read list_brand_design_guidance first and pass its exact content_hash. This preserves the section's assets and does not publish a Brand Release.
unmap_brand_interface_selection
ChatGPTPro owner only. Remove the interface intent mapping from one exact component variant using the current hash. Optionally pass the current intent as an additional match check. Preserves component source and unrelated variants. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
update_brand_custom_section
ChatGPTPro owner only. Update one exact custom Brand Page section. Supplied content replaces the selected fields. Read list_brand_page_sections first and pass its current hash. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
update_brand_interface_component
ChatGPTPro owner only. Update authored metadata of one exact Interface System component using its current hash. Supplied fields can overwrite metadata; captured Figma source and verified implementation evidence stay immutable. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
update_brand_interface_variant
ChatGPTPro owner only. Update or rename one exact authored component variant using the current Interface System hash. Supply a patch and/or new_variant. Supplied metadata can replace existing values; immutable imported source remains preserved. Does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
update_brand_page_draft
ChatGPTPro owner only. Save scoped Brand Page rules with the exact current snapshot_content_hash. item_guidance edits one color, logo, typography, asset or iconography snapshot ID. Iconography accepts style, stroke, weight, geometry, detail_level and a typed partial specification; nested fields merge and structured style fields stay synchronized. Assets accept usage_rule, including fonts. Existing IDs and unrelated metadata are preserved. Does not publish a Brand Release.
update_brand_page_item
ChatGPTPro owner only. Update one existing logo, font, typography, color or illustration record after the user's explicit request. No deletion or file replacement.
update_brand_snapshot_item
ChatGPTPro owner only. Update one exact snapshot item. Supplied fields can overwrite existing values. Read list_brand_snapshot_items first and pass its current snapshot hash. Preserves unrelated items; does not publish. Requires the exact active brand lock; existing ownership, OAuth, version and readback checks apply.
update_perture_settings
ChatGPTPro owner only. Change only agent behavior or Figma plugin sync booleans after the user's explicit request. Cannot edit billing, credentials, email or password.
upload_brand_file
ChatGPTPro owner only. Upload one logo, font, illustration, icon, pattern, imagery or asset from a local file after the user's explicit request. Send standard base64 bytes (maximum 8 MiB), a simple file_name and matching MIME type. For icon, pattern, imagery or asset, first read the draft and pass expected_snapshot_content_hash. For larger files use manage_brand_file_upload (up to 32 MiB in bounded chunks). Returns a file_id; registration and publication are separate. Never upload unrelated local files.
validate_external_output
ChatGPTCall once after an external host creates an output. Checks declared asset traceability and the complete source against the official brand context; a separate verify_brand_interface_source call is optional. For landing_page_hero, supply brand_interface_source with every planned semantic role, visual_evidence when reference-guided imagery is required, and complete generated source files. Include byte-identical local copies of official assets for verification. A generic HTML mockup cannot satisfy a required new illustration. Code-level findings and unverified artistic fidelity are reported separately.
validate_interface_system_usage
ChatGPTCall directly after generating component instances and before final UI verification. An active Interface System with eligible components requires at least one real usage; an empty manifest is a blocking error. Include intent and background/surface context for each instance when Brand Page mappings exist. Wrong component or variant for an intent, ambiguity, invalid registry IDs, properties, states or immutable implementation mappings block validation. The bundled local checker must bind exact artifact hashes to actual import/render source usage.
verify_brand_interface_source
ChatGPTCheck the complete generated HTML/JSX/TSX/CSS source against the same frontend contract used by Perture Build. Verify every planned role, immutable official component files and import/render use, tokens, typography, spacing, layout, assets, icon and textual rules. Reports source-level findings and uncovered visual/runtime evidence separately; does not inspect screenshots.
verify_generated_interface_source
ChatGPTLegacy single-control source check. Its source_verified status applies only to the requested action, never to the whole interface. Use verify_brand_interface_source to check a generated frontend output.