Skip to main content

Agentic Studio and the MCP Server

Use LinearB Catalog dashboard widgets from external AI agents through the LinearB MCP Server.

The LinearB MCP Server (Model Context Protocol) connects external AI clients, such as Claude and Cursor, to LinearB engineering data and capabilities.

The MCP Server uses the same tools as the Agentic Studio AI Assistant, allowing you to ask questions about engineering performance, retrieve metrics, and find and run widgets from LinearB Catalog dashboards without navigating the LinearB interface.

Important: MCP and the Agentic Studio AI Assistant can access widgets in LinearB Catalog dashboards. Personal dashboards and private or shared custom metrics created in Agentic Studio are not available through the MCP widget tools.

Company Admins control which roles can use Agentic Studio. These settings are separate from MCP Server authentication. Any LinearB user can connect to the MCP Server using OAuth without requiring Company Admin permissions solely for MCP access.

For more information about Agentic Studio permissions, see Agentic Studio Availability and Settings .

New to Agentic Studio? See Agentic Studio Overview .

What You Can Do Through the MCP Server

The MCP Server makes LinearB data accessible through natural-language requests in supported AI clients.

  • Search Catalog widgets: Find existing charts and KPIs that answer engineering performance questions.
  • Run Catalog widgets: Retrieve widget values or chart data for a specified team and reporting period.
  • Analyze engineering metrics: Explore delivery, quality, throughput, reliability, and supported AI impact metrics.
  • Explore teams and users: Look up teams, users, team iterations, and connected development tools.
  • Investigate engineering activity: Retrieve information about pull requests, issues, incidents, and related project management activity.

For a complete list of available tools and use cases, see MCP Server - Overview & Use Cases .

Find and Run LinearB Catalog Widgets

The MCP Server can retrieve existing widgets from LinearB Catalog dashboards.

Catalog dashboards contain predefined metrics and visualizations maintained by LinearB, including dashboards such as AI ROI, Engineering Health, and Software Factory.

Two MCP tools support this workflow:

Tool Description
find_widgets Searches LinearB Catalog dashboard widgets relevant to a question. Results include information such as the widget name, dashboard, description, and chart type.
run_widget Executes a selected widget, optionally using a specified start and end date, and returns the corresponding chart data or KPI value.

How It Works

  1. Connect an MCP-compatible AI client to the LinearB MCP Server.
  2. Ask a question about an engineering metric or dashboard widget.
  3. The client can use find_widgets to identify relevant Catalog widgets.
  4. The client can then use run_widget to retrieve the widget's data for the requested reporting period.
  5. Review the returned results, including charts, metric values, and any AI-generated interpretation.

Example Prompts

  • "Find a Catalog widget that shows rework rate."
  • "Run the Cost Per Effective PR widget for Q3 2026."
  • "Show the Rework Rate and PR Readiness widgets for the last 60 days."
  • "Find a widget that measures engineering delivery efficiency."
Tip: Include a specific team and reporting period in your question to help the AI client retrieve relevant results. Widget availability depends on the LinearB Catalog.

Supported Dashboard and Metric Access

Not all dashboards and custom metrics in Agentic Studio are available through MCP.

Data Type MCP Access
LinearB Catalog dashboard widgets Supported through find_widgets and run_widget.
Personal dashboards Not available through MCP widget tools.
Private or shared custom metrics created in Agentic Studio Not available through MCP widget tools.
Externally reported metrics submitted through the LinearB API Supported through the applicable get_reported_metrics and get_metrics tools.
Metrics Builder schema information Supported through get_db_schema for retrieving views, columns, and metric definitions. Custom SQL query execution is not supported.
Important distinction: Externally reported metrics submitted through the LinearB API are different from custom metrics created in Agentic Studio. Support for externally reported metrics does not mean that private or shared Metrics Builder widgets can be discovered or executed through MCP.

LinearB Catalog metrics are not exposed through the public LinearB API in the same way as MCP Catalog widget execution.

Metrics Builder Schema and SQL Queries

The get_db_schema tool provides information about the Metrics Builder data schema, including available views, columns, metric definitions, and supported derived metric calculations.

This information helps AI clients understand how engineering metrics are structured and calculated in LinearB.

SQL query support: The MCP Server can retrieve Metrics Builder schema information through get_db_schema, but it does not support executing custom SQL queries. The tool describes the available schema; it does not execute SQL statements or return the results of arbitrary SQL queries.

To retrieve supported engineering metrics, use get_metrics. To execute an existing LinearB Catalog widget, use run_widget.

Analyze Engineering and AI Impact Metrics

In addition to executing Catalog widgets, the MCP Server supports direct queries of engineering metrics through get_metrics.

These queries can be grouped or filtered by supported dimensions such as teams, repositories, users, time periods, and AI development tools.

Engineering Performance

Use natural-language questions to analyze metrics such as cycle time, pickup time, review time, time to approve, time to merge, and rework.

AI Adoption and Impact

Depending on the connected AI tools and available integration data, you can also explore AI usage and its relationship to engineering outcomes.

Supported analysis includes:

  • Comparing AI-assisted and manual development activity.
  • Distinguishing AI-involved work from work performed directly by AI.
  • Analyzing AI tool adoption and usage.
  • Examining AI-authored code, commits, and pull request activity.
  • Comparing engineering metrics across supported AI development tools.

Example Prompts

  • "Compare cycle time for AI-assisted and manual PRs over the last quarter."
  • "How many active Copilot and Cursor users did we have each week this quarter?"
  • "What percentage of our code changes last month were written directly by AI?"
Note: Available metrics vary by AI provider, integration method, and the data collected by LinearB. Not every AI tool supports every metric.

Example: Ask Claude for Cycle Time per Team

The following example demonstrates how an external AI client can query LinearB engineering metrics and summarize the results.

With Claude connected to the LinearB MCP Server, enter a question such as:

"Please provide me cycle time per team for DataBees from LinearB."

Claude uses available LinearB MCP tools to identify the team and retrieve its cycle time information.

The response can include a comparison of cycle time and its stages, including coding, pickup, review, and deploy time, for the selected team and related teams.

Claude displaying LinearB cycle time metrics for DataBees and related teams

The client can also summarize trends and identify differences between teams or reporting periods.

Claude summarizing weekly cycle time trends and engineering insights

This example illustrates how MCP can combine LinearB metrics with natural-language explanations to support engineering analysis.

Best practice: Always verify important numbers and calculations in LinearB before using AI-generated summaries for business decisions or sharing reports.

Authentication and Access

The LinearB MCP Server is generally available (GA). Any LinearB user can connect an MCP-compatible AI client using OAuth authentication.

OAuth is the recommended authentication method. It authenticates the user through their LinearB account and associates MCP requests with that user identity.

Company Admin permissions are not required solely to connect to the MCP Server.

Authentication Methods

  • OAuth (recommended): Authenticates through the user's LinearB account. Any LinearB user can use this connection method.
  • API key: The existing API-key connection method uses organization-level API-key authentication. OAuth is recommended for new connections.
Agentic Studio permissions: MCP Server authentication and Agentic Studio role permissions are separate. Company Admins control Agentic Studio access through its availability and settings controls. These controls should not be interpreted as a Company Admin requirement for connecting to the MCP Server.

The authentication method does not change the supported Catalog widget scope. Personal dashboards and private or shared custom metrics remain unavailable through the MCP widget tools.

To connect your AI client using the recommended OAuth authentication method, see Connect to LinearB MCP Server .

Additional Example Questions

Once connected, you can use the MCP Server to investigate other aspects of your engineering organization.

Planning and Sprints

  • "List Team Alpha's last three sprints."
  • "Show all high-priority bugs on the PLAT board that are still in progress."

Organization Setup

  • "Which of our integrations are currently failing?"
  • "How is change failure rate configured for our organization?"

Metrics and Benchmarks

  • "What externally reported metrics do we have in LinearB?"
  • "What does Review Depth measure?"
  • "How does our p75 cycle time compare with LinearB's industry benchmarks?"

Get Started

  1. Review MCP Server - Overview & Use Cases to explore available capabilities and tools.
  2. Follow Connect to LinearB MCP Server to configure your supported AI client using OAuth authentication.
  3. Ask a question about your engineering organization or a LinearB Catalog widget.
  4. Review the results and validate important values against LinearB.

Need help? Contact LinearB Support .

How did we do?

Agentic Studio Glossary

Troubleshooting Agentic Studio

Contact