MCP tools

The 226 tools an assistant can call when Klayson is connected from a connector directory: 83 read, 115 write and 28 sensitive writes.

Read tools only look. Write tools add to or change the connected workspace. Sensitive writes delete, overwrite, turn off or share something and may not be undoable: an assistant asks the person before calling one.

Connect

Klayson is a remote MCP server. Directory listings connect to https://api.klayson.com/mcp/directory; a custom connector can use https://api.klayson.com/mcp.

Sign-in is OAuth 2.1 with PKCE (S256) and dynamic client registration (RFC 7591). The server's 401 carries a WWW-Authenticate header pointing at its protected-resource metadata.

The person approves on Klayson's own page, signed in to the workspace the assistant will work in. The access token is a workspace key they can see and revoke in Settings.

Scopes: brand:read, brand:write, projects:read, projects:write, generate, references:read.

Read only: ticking Read only when approving gives a key with the read scopes alone. The server then lists and runs only the Read tools below.

Usage and limits

Every workspace has an amount of usage a month, set by its plan, that resets on the 1st. Making ads uses more of it than reading.

Tool calls are limited to 120 a minute and 20,000 a day for each workspace, and to a number of calls a month by plan.

Over a limit, the call returns a plain sentence saying which limit and when it opens again. Nothing is charged on top of the plan.

Ad search answers from Klayson's own ad index. The directory listings never read Meta, TikTok or Instagram live.

Errors

A tool that cannot do what it was asked returns isError: true with one sentence that says what went wrong and what to do next, for example get_frame_tree needs design_id. A tool this connection cannot use says it is not available here. A read-only connection asked for a write says it was connected read only.

Read tools (83)

get_me Get me
Who am I: user, workspace, brand summary, and this month's usage. Read-only: call it to verify the connection.
Inputs: none. Side effects: none, it only reads.

get_brand_extract Get brand extract
One brand read as Brand Extract v1: brand and facts, identity, materials, composition and page.
Inputs: read_id (string, required). Side effects: none, it only reads.

find_templates Find templates
The library templates that best fit what the person asked for, matched on their words.
Inputs: ask (string, required), limit, kit, type. Side effects: none, it only reads.

Write tools (115)

extract_brand Extract brand
Read a brand from its website and keep its materials: every stylesheet and the rendered page, the webfont files it serves, every logo candidate, its largest pictures and a screenshot.
Inputs: url (string, required), label, depth, fresh. Side effects: adds to or changes the connected workspace.

check_against_brand Check against brand
Check a picture or a canvas frame against the brand's own pieces: the palette, the ground, the case, where the logo sits, how large the product is.
Inputs: design_id, asset_id, image_url. Side effects: adds to or changes the connected workspace.

Sensitive writes (28)

scrape_brand Scrape brand
Scrape a website to (re)build the brand kit: colors, logo, about, voice. Replaces the scraped fields, keeps manual assets.
Inputs: url (string, required). Side effects: deletes, overwrites, turns off or sends something; it may not be undoable. Ask the person first. Starts background work; its result comes back from get_job.

Support