Created.ai

Showing 25 of 207 from the live probe.

• Login

  Authenticate as a specific user (POST /auth/login) and start a session scoped to this MCP connection. Backend-data tools (get_product, get_asset, list_assets) use this session's bearer token once logged in, falling back to a shared service account otherwise. Call set_context afterwards to select a team_id/brand_id. KEEP the returned `session_token`: every session-aware tool accepts it as an optional `session_token` argument — pass it on every call if your transport may reconn

• Login With Google

  Authenticate with a Google ID token (POST /auth/google) and start a session scoped to this MCP connection — the OAuth equivalent of `login`. Pass a Google-issued ID token (JWT) obtained from the client's Google sign-in; the backend verifies it against its GOOGLE_CLIENT_ID and returns the user's bearer token (the account must already have access/an invite). Backend-data tools then use this session's token. Call set_context afterwards to select a team_id/brand_id. KEEP the returned

• Login With Token

  Start a session from a backend access token the caller already holds, instead of signing in through this server. Where the token came from does not matter: a curl against POST /auth/login, the browser's color-tv-auth-access-token cookie, a secret in CI. This server cannot verify the token, so it calls GET /me once with it, sending the token as an explicit header and binding nothing until the backend answers. On an answer the session is bound to this connection and the user's

• Start Google Login

  Begin interactive Google sign-in against a REMOTE MCP server — step 1 of 2. Returns a `url` that redirects straight to Google's account chooser (like `gcloud auth login` — no intermediate page) and the one-time `state` for this attempt. When the server runs on the user's own machine (localhost MCP_PUBLIC_URL) the browser is opened automatically — check `browser_opened`; only if it is false, ask the user to open `url`. Immediately after this, call complete_google_login(state,

• Complete Google Login

  Finish interactive Google sign-in — step 2 of 2. Call immediately after start_google_login, with the `state` AND `poll_token` it returned: it WAITS (up to wait_seconds, default 60, max 300) for the user to finish signing in in the browser, then exchanges the captured Google ID token with the backend (POST /auth/google) and starts a session on this connection — same result as `login`. On timeout the state stays valid — just call this again. Errors immediately if the state expired

• Resume Session

  Reattach an existing login to the current MCP connection. Some clients (e.g. ChatGPT) reconnect the transport between calls, which loses the per-connection session even though the user already signed in — when a tool then reports "Not logged in", call this with the `session_token` returned by login / login_with_google / complete_google_login and retry the original call. Restores the full session including any team_id/brand_id set via set_context. NOTE: if the transport reconn

• Logout

  Log out the current connection's session. Invalidates the user's token server-side (POST /auth/logout — bumps credentialVersion, logging the user out on all devices), drops the local session so subsequent tools fall back to the shared service account, and invalidates the session_token so the login can no longer be resumed. No-op if not logged in (never logs out the shared service account). Returns {"status": "logged_out", "email": <email>} or {"status": "not_logged_in"}.

• Set Context

  Set or change the team_id / brand_id context for the current session (requires a prior login call on this connection). Pass just one field to switch team or brand without affecting the other. The selection stays isolated from other already-live connections that resumed the same token, while becoming the saved context for future reconnects.

• Preview Media

  Turn generated media urls into links the USER can actually open, and show the images. Pass the CDN urls you already have from any poll or get tool — `images[].url`, `videos[].url`, `savedVideos[]`, `results[].url`, `audioUrl`, `generatedVoiceUrl`, `imageUrl` — up to 12, in the order they should be shown. `labels` is an optional parallel list of captions and may be shorter than `media_urls`. Why this is needed: the media CDN is private. The urls the generation tools return a

• Me

  Fetch the current user's profile (GET /me). Uses the current session's bearer token when logged in, falling back to the shared service account otherwise. If the user signed in earlier but the client has since reconnected the transport, pass their session_token to this call (or resume_session first) — otherwise this describes the service account, not the user. The route also returns the whole image and video model catalogs and this tool drops them: they are 99.9% of the pa

• Estimate Generation Cost

  What a generation would cost in UFT before it runs (POST /billing/estimate). `mode` names the generation route; `params` is that route's request EXACTLY as it would be submitted to the matching generate_*/start_* tool — same camelCase field names, same asset refs — and is validated against that endpoint's own DTO, so a request the generation would reject comes back as a validation error naming the field, never as a number. team_id falls back to the session context (set_contex

• Get Team Balance

  The team's UFT balance and remaining tokens (GET /billing/balance/{teamId}). TEAM ADMINS ONLY — the backend answers 403 "Only team admins can view billing details" to every other member. That is not a failure of the flow: report it and carry on, since an estimate is meaningful on its own. team_id falls back to the session context (set_context). start_time/end_time (ISO 8601) turn the balance into the net change over that window (negative = spend) instead of the standing b

• List Teams

  List all teams the current user belongs to (GET /team), including each team's members and their roles/permissions. Uses the current session's bearer token when logged in, falling back to the shared service account otherwise. If the user signed in earlier but the client has since reconnected the transport, pass their session_token to this call (or resume_session first) — otherwise this lists the service account's teams, not the user's.

• List Brands

  List brands for a team (GET /brand), optionally filtered by a name search term. Uses the current session's bearer token when logged in, falling back to the shared service account otherwise. If the user signed in earlier but the client has since reconnected the transport, pass their session_token to this call (or resume_session first).

• Get Brand

  Fetch one brand by id (GET /brand/{id}) with its description, website, colours, fonts and typography classification. team_id falls back to the current session context (set_context). Returns the same fields as one list_brands entry — no team or products relations — so reach for this when you have an id, not to get more than the listing gave you. Errors when the brand is not in the resolved team (another team's brand and a deleted one both read that way) or when the user is not

• List Products

  List products for a team/brand (GET /product), paginated. team_id/brand_id fall back to the current session's context (set via set_context) when omitted; raises if neither is available. order is one of "ASC" | "DESC" | "AZ" | "ZA" | "MOST_POPULAR" (default "DESC"). Returns {"items": [...], "total": <int>}.

• Brain Index

  List live entries in the selected team/brand Brain, including each entry's slug, title, tags, folder path, and `space` (`brand` or `team`). Use this before reading or writing so you can reuse the correct slug/space and avoid duplicate entries. Brain tools require a real user login plus both team_id and brand_id in session context. They never fall back to the gateway service account, and neither tenant id is accepted as a tool argument.

• Brain Search

  Search the selected team/brand Brain. Results include the immutable slug and `space` needed by brain_read, brain_update, and brain_archive. `limit` must be 1-25. Search first before creating durable knowledge or when the index is too broad.

• Brain Read

  Read one live Brain entry by the slug and space returned by brain_index/brain_search. Returns its full Markdown content, metadata, outbound wikilinks, and backlinks. A missing, foreign, or inaccessible entry is deliberately reported the same way.

• Brain Create

  Create a durable Brain entry in the selected `brand` or `team` space. The core service derives an immutable slug from `title` and records MCP provenance. Search first: near-duplicate titles are soft-blocked with tenant-visible matches; use `confirm_new=true` only after reviewing those matches and confirming this is genuinely a separate topic. Markdown may contain [[wikilinks]].

• Brain Update

  Update one Brain entry in place, creating an auditable revision. Read it first, preserve useful existing content, and provide a non-empty one-line `change_note` saying what changed and why. `changes` must contain at least one of exactly: title, content, summary, tags, folder_id. Set summary or folder_id to null to clear it/move to the space root. Unknown keys are rejected. The slug remains immutable.

• Brain Archive

  Soft-archive one Brain entry and record a revision. Read it first and provide a non-empty one-line `change_note` explaining why it should leave the live Brain. This does not hard-delete history; a missing, foreign, or inaccessible entry is deliberately reported the same way.

• Get Product

  Fetch a product with its images, brand, colours, and font data from the backend.

• Get Asset

  Fetch an asset by ID from the backend. Raises an error if the asset is not found.

• List Assets

  Search and filter a team's asset library (GET /assets). Returns ONE page: {"assets": [...], "totalCount": <int>}. Page with skip/limit; do not fetch every page by default. team_id/brand_id fall back to the current session's context (set via set_context); team_id raises when neither is available. Because brand_id is taken from the context too, a brand must NOT be set in the context to list the whole team's library. Filters: - search — text query. AI semantic search, not