Skip to main content
🔌 Integration · Model Context Protocol

Bring your people & work context into Claude

Harmny’s MCP server lets Claude by Anthropic securely read your frameworks, career paths, reviews, org metrics, tasks, and projects — so answers are grounded in your real data, without leaving the chat.

Read-only · Scoped API keys · Included on Growth and above

What you can ask Claude

Every prompt maps to a real Harmny tool Claude can call. Nothing is made up — answers come from your live data.

“What career frameworks do we have?”

Claude lists every competency framework in your org.

get_frameworks

“List the competencies in the Engineering framework.”

Pull competencies for any framework, by name or id.

get_competencies

“What’s expected at L4 vs L5 in Product?”

Career levels and skill descriptions, side by side.

get_career_path

“Summarize Priya’s ratings, goals, and current level.”

A grounded snapshot for 1:1 prep or calibration.

get_employee_summary

“Give me a high-level snapshot of the org.”

Headline people metrics in one answer.

get_org_overview

“What urgent tasks are unassigned in Marketing?”

Filter tasks by project, assignee, status, or priority.

get_tasks

“List the active projects in the Growth space.”

Spaces and their live projects, optionally by space.

get_projects

Set it up in three steps

No engineering required. If you can copy and paste, you can connect Claude.

1

Generate a scoped MCP key

In Harmny, go to Settings → Intelligence → MCP Keys and create a key. Pick exactly the scopes you want Claude to read — frameworks, career paths, evaluations, people, or tasks.

2

Add Harmny to Claude

In Claude Desktop (or any MCP-compatible client), add Harmny as an MCP server and paste your key. Harmny publishes a discovery card at /.well-known/mcp/server-card.json so clients configure in seconds.

3

Ask Claude about your org

That’s it. Ask Claude anything about your people and work — answers come straight from live Harmny data, scoped to your organization and your key’s permissions.

Secure by default

Claude reads only what you allow, only within your organization.

Read-only by design

The Harmny MCP server exposes read tools only. Claude can look things up; it can never change, delete, or submit data back to Harmny.

Scoped API keys

Each key grants exactly the scopes you choose. A key without the “evaluations” scope can’t read employee summaries — the server refuses the tool.

Org-isolated

Every tool call is scoped to your organization at the database layer. Claude never sees another tenant’s data — the same multi-tenant guards that protect the Harmny API.

Rate-limited & audited

MCP traffic is rate-limited per key, and every tool call is logged with the key id and organization for a full audit trail.

Built on the open Model Context Protocol by Anthropic — the same standard Claude uses to connect to external tools.

Claude integration FAQ

Clarity. Motivation. Growth. For your whole team.

Free forever for up to 10 users. No credit card required. Full access from day one.

By signing up you agree to our Terms of Service and Privacy Policy.