Features

MCP Server

Use Koalr as a Model Context Protocol (MCP) server to give AI agents direct access to your engineering metrics.

The Koalr MCP server exposes your engineering data as Model Context Protocol tools. This lets AI agents — Claude, Cursor, Windsurf, or any MCP-compatible client — query live metrics from your organization without you having to copy-paste dashboards.

What it enables

  • Ask Claude "which repos are at the highest deploy risk right now?" and get a real answer backed by live data
  • Build AI workflows that automatically flag risky PRs, summarize DORA trends, or escalate coverage regressions
  • Connect Cursor or Claude Desktop to Koalr so your IDE assistant has context about your delivery metrics

Installation

Option A — Remote (no install required)

Connect to Koalr's hosted MCP server at mcp.koalr.com. This works in Smithery, Glama, or any client that supports Streamable HTTP transport.

Remote URL:

https://mcp.koalr.com/{your_api_key}/mcp

Replace {your_api_key} with your Koalr API key.

Or discover Koalr on registry platforms:

  • Smithery — one-click connect with automatic auth
  • Glama — connect and inspect every tool call

Option B — Local stdio install

Claude Code

claude mcp add koalr \
  -e KOALR_API_KEY=your_api_key \
  -- npx -y koalr-mcp@latest

Claude Desktop

Add the Koalr server to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "koalr": {
      "command": "npx",
      "args": ["-y", "koalr-mcp@latest"],
      "env": {
        "KOALR_API_KEY": "your-api-key-here"
      }
    }
  }
}

Restart Claude Desktop. The Koalr tools will appear as available tools.

Cursor

Add to .cursor/mcp.json in your project (or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "koalr": {
      "command": "npx",
      "args": ["-y", "koalr-mcp@latest"],
      "env": {
        "KOALR_API_KEY": "your-api-key-here"
      }
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "koalr": {
      "command": "npx",
      "args": ["-y", "koalr-mcp@latest"],
      "env": {
        "KOALR_API_KEY": "your-api-key-here"
      }
    }
  }
}

Get an API key

  1. Go to Settings → API Keys in the Koalr dashboard
  2. Click Create API Key
  3. Give it a name like "Claude Desktop" or "Cursor"
  4. Copy the key — it's shown once

Available tools

20 tools across 8 categories — all backed by Koalr's live production API.

Deploy Risk

ToolDescription
score_pr_for_deploy_riskScore a pull request 0–100 using 36 research-validated signals: change entropy, DDL migration detection, author file expertise, PR size, CODEOWNERS violations, deploy timing, and more

Example:

"How risky is PR #247 before we merge for the release?"


DORA & Delivery

ToolDescription
get_dora_summaryDORA metrics (deploy frequency, lead time, change failure rate, MTTR) for the org or a team
get_dora_trendWeek-over-week trend for any single DORA metric

Example:

"What's our current DORA tier and how does lead time compare to last quarter?"


Pull Requests

ToolDescription
get_pr_summaryPR metrics: cycle time, throughput, review health, PR size distribution
get_open_prsCurrently open PRs with age, size, and reviewer assignments
get_at_risk_prsOpen PRs that are stale (>3 days), large (>500 lines), or unreviewed

Example:

"Are there any PRs that have been sitting open for more than a week?"


Teams & People

ToolDescription
list_teamsAll teams with member counts and recent activity
get_teamDORA + PR metrics for a specific team
list_team_membersMembers of a specific team with contribution signals
get_developerSummary metrics for a specific developer — throughput, cycle time, review participation
list_top_contributorsTop contributors ranked by activity signals

Example:

"Which developers have the fastest cycle times on the platform team?"


Organization

ToolDescription
get_org_healthComprehensive org snapshot: DORA tier, cycle time percentiles, active teams, incident rate
get_well_being_summaryTeam well-being: focus time, meeting load, context switching, burnout risk

Example:

"Give me a full health check of the engineering org this quarter"


Repositories & Coverage

ToolDescription
list_repositoriesAll connected repositories
get_repositoryActivity, PR throughput, coverage, and deploy risk for a specific repo
get_coverage_summaryOverall coverage %, trend, and test pass rate across connected repos

Example:

"Which repositories dropped coverage this week?"


Incidents & AI Adoption

ToolDescription
list_recent_incidentsRecent incidents from PagerDuty/OpsGenie with severity and MTTR
get_ai_adoption_summaryGitHub Copilot and Cursor adoption rates, acceptance rate, estimated hours saved
get_ai_adoption_trendAI tool adoption trend over time

Example:

"How many P1 incidents did we have last month and what were the affected services?"


ToolDescription
searchSearch across developers, PRs, issues, repositories, and reports

Self-hosted / API key scoping

The KOALR_API_KEY must belong to a user with at least Member access. For read-only automation, create a dedicated service account.

To point the MCP server at a self-hosted or staging Koalr instance, set KOALR_API_URL:

{
  "env": {
    "KOALR_API_KEY": "your-key",
    "KOALR_API_URL": "https://api.your-koalr-instance.com"
  }
}

Example prompts

Once connected, you can ask natural-language questions directly:

  • "What's our average PR cycle time and how does it compare to last month?"
  • "Which PRs should I be worried about before Friday's release?"
  • "We had an incident last night — was there a high-risk deploy in the 4 hours before?"
  • "Which team has improved their DORA tier the most in the last 90 days?"
  • "Are there any repos with coverage below 40% that had changes merged this week?"
  • "Score this PR for deployment risk" (then paste the PR diff or description)
  • "Which developers have the highest review participation rate?"
  • "How is our AI coding tool adoption trending across teams?"

The MCP server translates these questions into tool calls, fetches live data from your Koalr organization, and returns structured answers — no copy-pasting dashboard screenshots required.