To let an assistant use these tools, grant AI agents on the consent screen, or just
Agent access (WebMCP, A2A, ChatGPT and Claude apps) under its Fine-tune. The scope is agent_access.
Read tools work with Read only or Read & write access; write tools only with
Read & write.
Your assistant doesn't see these tools straight away. It finds them with search_tools or
describe_tools, and runs one with use_tool, passing its name and
arguments. Most assistants do this for you: just ask for what you want.
What your assistant is told about this group
The ways other people's AI agents can reach your agent. WebMCP: tools for AI assistants in visitors' browsers on your web pages. A2A agent cards: other businesses' AI agents ask your agent questions, server to server. ChatGPT and Claude apps: public connector URLs anyone can add to ChatGPT or Claude. get_agent_access_insights shows what those agents did. Creating or enabling any of them makes something public: confirm with the user first. An app also lets anyone with its connector URL read, word for word, the context blocks on its agent's live version: tell the user. Deleting one, or rotating a WebMCP key, breaks it for everyone using it. Needs a manager or owner, on a plan with agent access.
The descriptions below are the ones your assistant reads, so they're written for it. When a connection covers more than one
organisation, the assistant passes org (the organisation's slug, from list_orgs) to say which one.
get_agent_access_overview
Read Get agent access overview
How many WebMCP integrations, A2A agent cards, ChatGPT apps and Claude apps the org has. Manager or owner, on a plan with agent access.
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_agent_access_insights
Read Get agent access insights
What outside AI agents did with the org (WebMCP tool calls on its pages, A2A calls, ChatGPT and Claude app calls): totals against the prior window, questions answered, leads and callbacks captured, success rate, breakdowns by tool, outcome, surface and page, daily counts, top questions and recent events. period: today, week, month (so far), last7, last30, last90, last_month (last30 by default), or custom with from/to (YYYY-MM-DD). Manager or owner.
Arguments
| Field | Type | Description |
|---|---|---|
| period | string |
One of: "today", "week", "month", "last7", "last30", "last90", "last_month", "custom"
|
| from | string | |
| to | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_webmcp_integrations
Read List WebMCP integrations
The org's WebMCP integrations: embed placements that expose tools to AI agents browsing the org's web pages, with their allowed origins and switched-on tools.
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_webmcp_integration
Read Get a WebMCP integration
One WebMCP integration with the full tool catalogue: which tools are on, each tool's config fields and current settings. The publishable key is not shown.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
create_webmcp_integration
Write Create a WebMCP integration
Creates a WebMCP integration, enabled at once: once its snippet is on the listed origins, outside AI agents browsing those pages can call its tools. Confirm with the user first. enabled_tools defaults to every catalogue tool.
Arguments
| Field | Type | Description |
|---|---|---|
| name required | string | |
| allowed_origins | array of string | Websites the embed runs on, e.g. https://www.example.com (a page address is cut to its origin). |
| enabled_tools | array of string | Catalogue tool names (see get_webmcp_integration). |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_webmcp_integration
Write Update a WebMCP integration
Changes only the fields given. Enabling it, adding origins or switching tools on exposes the org's agent to outside AI agents on those pages: confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| name | string | |
| enabled | boolean | |
| allowed_origins | array of string | Replaces the whole list. Websites like https://www.example.com; a page address is cut to its origin. |
| enabled_tools | array of string | Replaces the whole list. |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
delete_webmcp_integration
Write Delete a WebMCP integration
PERMANENT. Deletes the integration: its key stops working, so the embed on live pages stops exposing tools. Confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
rotate_webmcp_key
Write Rotate a WebMCP key
Mints a new publishable key and returns it once. The old key stops working immediately, breaking the live embed until the snippet is redeployed: confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
configure_webmcp_tool
Write Configure a WebMCP tool
Sets one tool's options on a WebMCP integration, using the keys from that tool's config_fields (get_webmcp_integration). Replaces the tool's config; an empty config clears it.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | The integration id. |
| tool_name required | string | |
| config | object | config_fields key → value (a list for multiselect fields). |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_a2a_integrations
Read List A2A agent cards
The org's A2A (Agent2Agent) integrations: agent cards that let other AI agents call the org's agent server to server, with where each card is hosted.
Arguments
| Field | Type | Description |
|---|---|---|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_a2a_integration
Read Get an A2A agent card
One A2A integration: its answering agent, hosting site (or own-server download), public card URL and the endpoint other agents call.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
create_a2a_integration
Write Create an A2A agent card
Creates an A2A integration, enabled by default: other AI agents can then call the org's agent through it (publicly, if assigned to a hosted site). Confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| name required | string | |
| description | string | What the card tells other agents about the org (600 chars max). |
| site_id | string | A hosted site to serve the card from; blank = the org hosts it on its own server. |
| agent_id | string | The live agent that answers; blank = the org's first live agent. |
| enabled | boolean | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_a2a_integration
Write Update an A2A agent card
Changes only the fields given. Enabling it or assigning a site publishes the org's agent to outside AI agents: confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| name | string | |
| description | string | What the card tells other agents about the org (600 chars max). |
| site_id | string | A hosted site to serve the card from; blank = the org hosts it on its own server. |
| agent_id | string | The live agent that answers; blank = the org's first live agent. |
| enabled | boolean | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
delete_a2a_integration
Write Delete an A2A agent card
PERMANENT. Deletes the integration: its card and endpoint stop answering, breaking any agent that uses them. Confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
list_assistant_apps
Read List ChatGPT and Claude apps
The org's assistant apps: public MCP endpoints people connect to from ChatGPT or Claude. platform narrows to chatgpt or claude.
Arguments
| Field | Type | Description |
|---|---|---|
| platform | string |
One of: "chatgpt", "claude"
|
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
get_assistant_app
Read Get a ChatGPT or Claude app
One assistant app: connector URL, answering agent, the tools this org can switch on (with each tool's config fields and settings) and the plugin package metadata.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
create_assistant_app
Write Create a ChatGPT or Claude app
Creates an assistant app, enabled by default: anyone with its connector URL can use the org's tools from ChatGPT or Claude, anonymously. They can also read, word for word, every context block on the live version of the agent behind the app, plus that agent's persona and the app's description. Before creating it, tell the user this, suggest they review those context blocks (get_agent_version shows them), and get their go-ahead. enabled_tools defaults to every tool the org can have.
Arguments
| Field | Type | Description |
|---|---|---|
| platform required | string |
One of: "chatgpt", "claude"
|
| name required | string | |
| description | string | Instructions for the host about the app (2000 chars max). |
| agent_id | string | The live agent behind the app; blank = the org's first live agent. |
| enabled | boolean | |
| enabled_tools | array of string | Tool names (see get_assistant_app). |
| widget_theme | object | Quote card colours: key → #rrggbb (highlight, link, button_bg, button_fg, border, surface, mark, success, warning). |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_assistant_app
Write Update a ChatGPT or Claude app
Changes only the fields given. Enabling the app or switching tools on exposes them to anyone using its connector URL: confirm with the user first. Changing agent_id changes what the public can read: anyone with the URL can read, word for word, every context block on the new agent's live version, plus its persona. Say so, and suggest the user reviews those blocks (get_agent_version) first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| name | string | |
| description | string | Instructions for the host about the app (2000 chars max). |
| agent_id | string | The live agent behind the app; blank = the org's first live agent. |
| enabled | boolean | |
| enabled_tools | array of string | Tool names (see get_assistant_app). |
| widget_theme | object | Quote card colours: key → #rrggbb (highlight, link, button_bg, button_fg, border, surface, mark, success, warning). |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
delete_assistant_app
Write Delete a ChatGPT or Claude app
PERMANENT. Deletes the app: its connector URL stops answering, breaking every ChatGPT or Claude install that uses it. Confirm with the user first.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
update_assistant_app_plugin
Write Update an app's plugin package
Edits the downloadable plugin package metadata (listing text, prompts, brand colour, SKILL.md, widget domains). Only the fields given change; the logo is set in the dashboard.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | |
| config required | object | Any of: plugin_name, version, display_name, short_description, long_description, developer_name, category, website_url, privacy_policy_url, terms_of_service_url, brand_color, skill_name, skill_description, skill_body, widget_domain, capabilities, default_prompts, keywords, widget_connect_domains, widget_resource_domains. capabilities, default_prompts, keywords, widget_connect_domains, widget_resource_domains are lists of strings. |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |
configure_assistant_app_tool
Write Configure a ChatGPT or Claude app tool
Sets one tool's options on an assistant app (products a quote tool prices, callback calendar, CRM endpoint), using the keys from its config_fields (get_assistant_app). Replaces the tool's config; an empty config clears it.
Arguments
| Field | Type | Description |
|---|---|---|
| id required | string | The app id. |
| tool_name required | string | |
| config | object | config_fields key → value (a list for multiselect fields). |
| org | string | Organisation slug: required when this authorization covers more than one organisation (see list_orgs) |