Leapsome MCP (Model Context Protocol)

The Leapsome MCP (Model Context Protocol) integration connects approved AI clients, such as Claude, Notion AI, and Cursor, directly to Leapsome. It gives AI agents a focused, permission-controlled way to retrieve Leapsome context and, where explicitly enabled, take selected actions on your behalf.

Rather than replicating every Leapsome screen or API endpoint, the MCP is designed around outcomes. Describe what you want to accomplish, and let your AI client discover and combine the capabilities available to you.

If your AI client connects to other services such as Slack, Notion, or a calendar, it can combine their context with Leapsome data in the same conversation. Each connected service continues to enforce its own access controls.

Table of contents

• Enabling MCP in Settings
• Connecting your AI client
• What you can do with the Leapsome MCP
• Important notes

When to use this feature

Use the Leapsome MCP when you want to bring Leapsome data and actions into an AI client you already use. Common use cases include:

• Preparing for conversations. Build a structured 1:1 briefing from meeting context, talking points, notes, and action items, then let your agent add relevant goals, recent feedback, or calendar events from connected services.
• Drafting reviews with richer context. Combine review, goal, and feedback information with documents from other connected tools to prepare draft responses. Drafts can be reviewed and saved after confirmation; MCP does not submit or finalize them.
• Analyzing people insights. Bring structured Leapsome analytics into your AI client to visualize, explain, or combine with other datasets.
• Taking lightweight actions. Give praise, create or update personal goals, add a meeting talking point, or request time off, all from within your AI client.
• Building broader workflows. Let an agent combine Leapsome reads and confirmed actions with other tools your organization uses.

Click here to watch the video that walks you through enabling the MCP, connecting via Claude, and example use cases. 

Enabling MCP in Settings as an Admin

1. Go to 'Settings' > 'Integrations and imports' > 'MCP (AI connectors, Beta)'.
2. Tick the box 'Enable the MCP for this organization' to enable it.
3. Scroll down and select the read scopes your organization wants to permit.
4. Optionally, enable write actions by area (Goals, Meetings, Feedback, Absence, Workflows, Reviews). Write access is off by default and must be explicitly enabled per area.

After enabling MCP, you can also grant access to the relevant user groups, if you wish. To do so:

1. Go to 'Settings' > 'Employees' > 'Feature access matrix' > 'AI Features'.
2. Enable 'Use the MCP integration' for the applicable user groups.

Users who do not have this permission enabled will not be able to connect their AI client to Leapsome.

Connecting your AI client

Connecting via Claude (preconfigured connector)

1. Open Claude and navigate to Connectors → Leapsome.
2. Sign in with your Leapsome account.
3. Review and approve the scopes you want to grant.

All read scopes are pre-selected. Write actions and Payroll data access require explicit selection.

Connecting via another MCP-compatible client

1. In Leapsome, go to Settings > Integrations and imports > MCP (AI connectors, Beta).
2. Copy your MCP server URL: https://www.leapsome.com/mcp.
3. Add the URL manually in your AI client's MCP configuration.
4. If your client does not support OAuth, use a personal access token instead, available in the same settings page.

Please note: most MCP clients support only one connection to the same MCP server at a time. If you use multiple Leapsome accounts, disconnect and re-authenticate with the correct account when switching.

What you can do with the Leapsome MCP

The data and actions available to you depend on your enabled Leapsome modules, the scopes your organization has approved, and your existing Leapsome permissions.

Read capabilities

• People and organization: Employee profiles, org chart, teams, locations, birthdays, work anniversaries, skills and competency framework, and visible custom attributes. Also includes skill matrix and competency data, and custom attribute definitions with their selectable values.
• Knowledge base: Search over your company knowledge base and the global HR knowledge base. Only content you already have access to in Leapsome is returned.
• Goals: Goals and cycles visible to you, including comments, update feed, and cycle statistics.
• Meetings: 1:1 and team meetings, agendas, action items, shared and private notes, meeting analytics, and AI-generated summaries where available. Audio files and transcript text are not exposed.
• Feedback: Instant feedback and praise items visible to you, and feedback volume analytics.
• Reviews: Review cycles, participants, participant statuses, your review tasks, and cycle details you can already see in Leapsome.
• Surveys: Survey results and comments visible to you. Anonymity protections are enforced on the server side.
• Learning: Learning paths, your progress, path statistics, enrollment lists where permitted, and accessible learning elements.
• Absence: Absences and time-off requests visible to you, absence balances, policies, and holiday calendars.
• Analytics: Analytics dashboards, widgets, reports, and time-tracking analytics you already have access to.
• Forms: Forms assigned to you, participant-facing form details, and your own form submissions. Tenant-wide form administration, submission detail, and change logs are not exposed.
• Time tracking: Your timesheets, your assigned time-tracking policy, and the projects and project clients you can log time against.
• Approvals and workflows: Approval flows you initiated, approve, or are the subject of, including step status, approvers, and decisions. Self-service workflow templates visible to you.
• Payroll (sensitive; explicit opt-in required): Payroll policies, cycles, participant data, salary details, and payroll analytics you can already access in Leapsome.

Write actions

Write actions are off by default. Your organization must enable them per area, and you must explicitly grant access to each area when connecting your AI client. Every action uses a proposal-and-confirmation step; nothing is saved until you approve it.

• Goals: Create personal goals, edit common fields, update progress and progress status, and comment on goals you can access.
• Meetings: Add talking points to meetings you can edit. Creating action items, commenting on meeting items, and accepting or rejecting AI suggestions are not exposed.
• Feedback: Send named instant feedback or praise to a single recipient. Feedback comments, reactions, anonymous suggestions, and Q&A board questions are not exposed.
• Absence: Request time off for yourself. Updating or cancelling absence requests and changing out-of-office substitutes are not exposed.
• Workflows: Trigger self-service workflows you are eligible for. Acting on workflow step assignments is not exposed.
• Reviews: Create or update draft responses you can already edit. This does not submit, share, or finalize a review.

Important notes

• The Leapsome MCP is currently in beta. Capabilities will expand based on real usage and unmet requests.
• It is a focused, agent-first interface, not a replacement for Leapsome's APIs or a full copy of the platform.
• Leapsome only exposes data and actions you can already access within the platform. All existing visibility, permission, and anonymity rules apply.
• Text custom attributes are not fully supported in this version.
• Other connected services in your AI client retain their own access controls. Combining their data with Leapsome happens inside your AI client, not within Leapsome itself.