MCP

Svit.One provides a remote Model Context Protocol (MCP) server. Compatible AI assistants and applications can use it to list your projects, work with tasks, people, and opportunities, read or create comments, list sites, and list or trigger site deployments.

Connect an MCP client

Add the following remote MCP server URL to your client:

https://api-external.svit.one/mcp

Select Streamable HTTP as the transport if your client asks for one. Configuration field names vary by client, but a typical remote-server entry looks like this:

{
    "mcpServers": {
        "svit": {
            "url": "https://api-external.svit.one/mcp"
        }
    }
}

When prompted, sign in to Svit.One and approve access. The client can access only projects you belong to.

How to use the tools

Start with list_projects, then use its projectId with the other tools. Useful discovery flows include:

• Call list_folders before filtering tasks by folder or creating or updating a task in a folder.
• Call list_project_users before filtering tasks by assignee, assigning a task, or setting an opportunity or person owner.
• Call find_person to look up a contact by email, phone, or id. Use get_person when you already have a person id and a missing contact should fail.
• Call list_tasks before reading or creating comments on a task.
• Call list_sites before list_deployments or trigger_deployment.
• Call list_opportunities or find_opportunity before reading or updating an opportunity.

Project tools

list_projects

Lists every Svit.One project the authenticated user belongs to.

Inputs: None.

Output: A projects array whose entries contain:

list_tasks

Lists tasks in a project, with pagination, search, filtering, and sorting.

Output: A tasks array of task objects and a pagination object containing total, page, limit, and totalPages numbers.

create_task

Creates a task in a project.

Output: The created task object.

update_task

Updates an existing task. Only provided fields change. Pass null to clear a nullable field or an empty assigneeIds array to remove all assignees. At least one field to update is required. The connected user needs tasks:edit access in the project.

Output: The updated task object.

list_folders

Lists task folders in a project.

Output: A folders array whose entries contain:

list_project_users

Lists the active members of a project.

Output: A users array of user objects.

Site tools

list_sites

Lists sites in a project. Use a site's id with list_deployments or trigger_deployment. Those tools only work when deploymentVersion is 3.

Output: A sites array whose entries contain:

list_deployments

Lists deployments for a site. The site must belong to the project and use deployment version 3. Legacy sites are rejected.

Output: A deployments array of deployment objects.

trigger_deployment

Schedules a deployment for a v3 site (deploymentVersion 3) from the site's latest artifacts. This is the MCP equivalent of Publish in the site editor: it duplicates current artifacts, creates a deployment with status SCHEDULED, and enqueues it for the worker. It does not use the legacy immediate-publish path. Sites that are not on deployment version 3 are rejected.

Output: The scheduled deployment object. Status is SCHEDULED.

Opportunity tools

list_opportunities

Lists opportunities in a project, with pagination, name search, and filters. Results follow the authenticated user's CRM opportunity and pipeline view permissions.

Output: An opportunities array of opportunity objects and a pagination object containing total, page, limit, and totalPages numbers.

find_opportunity

Looks up one opportunity without erroring when nothing matches. Provide at least one of opportunityId, personId, or name. Name matching is exact and case-insensitive. This is a soft lookup: a miss returns found: false and opportunity: null instead of an error. read_opportunity still throws when the ID is missing or inaccessible.

Output: { found, opportunity } where opportunity is an opportunity object or null.

read_opportunity

Reads an opportunity by ID. The authenticated user must have access to both CRM opportunities and the opportunity's pipeline.

Output: An opportunity object.

create_opportunity

Creates an opportunity in a project's default pipeline and initial stage.

Output: The created opportunity object.

update_opportunity

Updates an existing opportunity. Only provided fields change. Pass null to clear the description or owner. At least one field to update is required. The connected user needs edit access to the opportunity's pipeline.

Output: The updated opportunity object.

Person tools

Start with list_projects, then call these tools with that projectId. Use list_project_users to resolve ownerUserId values. find_person is a soft lookup: a miss returns { found: false, person: null }. get_person fails when the person does not exist.

find_person

Looks up a person by id, email, or phone. At least one of personId, email, or phone is required. A miss is not an error.

Output:

get_person

Loads a person by ID. A missing person is an error. Use find_person when a miss should not fail.

Output: A person object.

update_person

Updates an existing person. Only provided fields change. Pass null to clear nullable description, organizationId, ownerUserId, or position. At least one field to update is required. The connected user needs contacts:edit access in the project.

Output: The updated person object.

create_person

Creates a person in a project. The connected user needs contacts:create access.

Output: The created person object.

Comment tools

Comments are polymorphic: they can be attached to different kinds of Svit.One objects. For task comments, use task as the objectType and a task ID as the objectId.

Supported objectType values are dataset, site, project, task, and conversation.

list_comments

Lists comments attached to an object in a project.

Output: A comments array of comment objects.

create_comment

Adds a comment to an object in a project.

Output: The created comment object.

Shared result objects

Opportunity object

Person object

Task object

User object

Comment object

Deployment object

Task values

Task statuses are todo, in-progress, review, and done.

Task priorities are high, medium, low, and none.