The agent itself
Space MCP tools in the The agent itself scope (self). Reach these with a wks_live_ key over space-mcp or space-api.
Tools in the self scope, callable over space-mcp (tools/call) or space-api (REST) with a wks_live_* key that grants self.
complete_setup_item
Scope: self:write
Mark one of your setup items done (or not done) after you did it or verified it yourself — say which by its key from my_setup. The owner's page shows the tick with your name on it. Never mark something done that you did not check.
| Parameter | Type | Required | Description |
|---|---|---|---|
key | string | yes | From my_setup |
done | boolean | Default true |
get_my_instructions
Scope: self:read
Your standing instructions exactly as stored — the brief your owner wrote (and you may have changed), with its @-tags in the chat's token grammar (@integrationLabel). Read this before update_my_instructions so you send the whole text back with every tag you are not changing kept as it is. The copy you read at the top of a session is the same text with the tags resolved to handles.
my_setup
Scope: self:read
Your setup list, if you were made from a recipe: each thing you need from the owner before your first job, and whether it is done — the platform checks documents and app records itself; a manual item is done when someone marked it. Empty when you have no such list.
set_my_profile
Scope: self:write
Change your own name or picture — when the owner asks you to, or when you were made with a placeholder. Send only what changes. The picture is a URL of an image already on the web or in this space's assets.
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Your new display name, 2–60 characters | |
avatar_url | string | A URL to your new picture; an empty string removes it |
update_my_instructions
Scope: self:instructions
Rewrite your own standing instructions — the brief you read at the top of every session. Use it for something that should HOLD from now on (a rule the owner gave you in chat, a way of working you settled on), never for a one-off. Send the WHOLE new text (up to 8000 characters): call get_my_instructions first and keep everything you are not changing, @-tags included — you may keep tags in the resolved form you read them in (@Leads [table:<id>]) or the token grammar (@integrationLeads). Every change is kept in History and shown to your owner as a line in this chat; when your owner asks first, the change waits on a card here — say briefly what you want to change and why, and wait.
| Parameter | Type | Required | Description |
|---|---|---|---|
instructions | string | yes | The full new text. Empty clears your instructions. |
reason | string | One line on why — shown with the change. |