# Tools reference

> Every tool the Lorito MCP server exposes, with its arguments and what it returns.

Source: https://www.lorito.ai/docs/integrations/mcp-tools

These are the tools available over the [MCP server](https://www.lorito.ai/docs/integrations/mcp). Read-only tools return data as JSON. Tools that make changes confirm what they did. Every call runs with your account's permissions, so you can only act on organisations and projects you belong to.

IDs are UUIDs. You typically start from `project_list` to find a project, then pass its ID to the other tools.

## Projects

| Tool             | What it does                                        | Arguments                                                                       |
| ---------------- | --------------------------------------------------- | ------------------------------------------------------------------------------- |
| `project_list`   | List the projects in an organisation, newest first. | `org_id` (required)                                                             |
| `project_get`    | Get a single project's details.                     | `project_id` (required)                                                         |
| `project_create` | Create a new project in an organisation.            | `org_id` (required), `name` (required), `client_name` (optional)                |
| `project_update` | Update a project's name, client name, or status.    | `project_id` (required), `name`, `client_name`, `status` (`active` or `paused`) |

`project_list` and `project_get` return fields such as `id`, `name`, `client_name`, `status`, and `created_at`. Asking for an organisation you do not belong to returns an empty list rather than an error.

## Knowledge sources

| Tool            | What it does                                               | Arguments                                                                                                                                                                    |
| --------------- | ---------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `source_list`   | List a project's knowledge sources, newest first.          | `project_id` (required)                                                                                                                                                      |
| `source_add`    | Add a website as a knowledge source and start indexing it. | `project_id` (required), `url` (required), `label` (optional, defaults to the site's hostname), `page_limit` (optional, default 50, capped at your plan's pages per website) |
| `source_delete` | Delete a knowledge source and all of its indexed content.  | `source_id` (required)                                                                                                                                                       |

> **Warning**
>
> `source_add` begins indexing immediately, and `source_delete` permanently removes the source along with everything indexed from it. Deleting cannot be undone.

`source_add` discovers the site's pages (up to `page_limit`) and queues them for indexing, then returns the new source's ID. `source_list` returns fields such as `id`, `url`, `label`, `status`, and `pages_found`.

## Agents

| Tool           | What it does                                                                                                | Arguments                                                                                   |
| -------------- | ----------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------- |
| `agent_get`    | Get the default agent's configuration for a project.                                                        | `project_id` (required)                                                                     |
| `agent_update` | Update the agent's persona, model, instructions, or knowledge search. Only the fields you send are changed. | `agent_id` (required), `persona_name`, `model`, `system_prompt`, `search_knowledge_enabled` |

`agent_get` returns the agent's `id`, `persona_name`, `persona_avatar`, `model`, `system_prompt`, and `search_knowledge_enabled`. Pass that `id` to `agent_update`.

> Updating `system_prompt` saves a new prompt version, so the change is tracked alongside edits you make in the dashboard. For `model`, use a model ID such as `anthropic/claude-sonnet-5.5`, or `recommended` to follow the latest recommended model.
