LinearB MCP server: ask about cycle time and DORA metrics

What the hosted server reaches, how each client connects to it, and why every metric it returns needs a time window and a team.

Checked 3 October 2026 · from LinearB's help-centre articles on the MCP server, an archived March 2026 copy, and the live endpoint's OAuth metadata

One of 91 guides to monitoring and developer tools. Each covers what an AI coding agent can get out of the tool through MCP, the protocol that lets the agent query the tool directly.

What is the LinearB MCP server?

LinearB is an engineering-intelligence platform. It reads from Git providers, issue trackers and CI to report cycle time, throughput, review latency and AI tool usage.

MCP (the Model Context Protocol) is the standard AI clients use to call outside tools. With LinearB connected, an assistant can ask how long reviews took last sprint without anyone opening a LinearB dashboard.

The help centre has documented the server since 17 September 2025, when the "Exploring the Power of LinearB's MCP Server" article was published. It was labelled Beta then and is generally available now.

Where LinearB documents and serves the MCP server:

A hosted server has no package or repository to install. The help centre at linearb.helpdocs.io is where setup lives.

LinearB measures delivery, so the server reaches code review and release data and stops before production telemetry.

The strongest use is a manager's question about a team over a period: cycle time, merge rate, release contents. The server reports on engineering work; it does not observe running software.

What can an AI agent do with LinearB?

LinearB's current docs describe data domains rather than tool names. The help-centre article as captured on 7 March 2026 listed twelve tools: metrics, repositories, contributors, releases, incidents, issues, services, users, teams, branches, pull_requests and pm_entities.

Can LinearB watch your own AI agents?

LinearB's AI Analytics reports on engineers' use of AI coding tools. It does not record the calls your own agent makes, so most agent-observability signals fall outside it.

LinearB MCP server limits: where answers come back incomplete

Each limit below comes from the server answering with aggregates over people and time.

How to configure LinearB MCP for agents

1. 1

   Use OAuth where the client supports it

   Access then follows the signed-in user's LinearB role, and no key sits in a config file.

2. 2

   Treat an API key as organization-wide

   LinearB's examples name it LINEARB_ADMIN_API_KEY. Store it as a secret; LinearB documents a Kubernetes Secrets pattern for this.

3. 3

   Decide who can query people metrics

   Contributor-level cycle time and review data describe individuals, so the user role matters.

4. 4

   Check headline figures in the dashboard

   LinearB's own guidance says to verify important results before official reporting.

Paste this into your agent’s instructions

Every LinearB metric is an aggregate over a period and a set of people. The block below makes the agent state both, so a figure cannot travel without its scope.

When you answer using LinearB MCP tools, state:
- The tool you called and the metric it returned.
- The date range and the teams or repositories the metric
  covers.
- Any filter you applied, such as a team, contributor or
  service.
- That engineering metrics describe a period, so a figure
  without its window means nothing.
Never state that data does not exist. State the tool you
called and what it returned.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

LinearB's dashboards remain the place to verify figures, and are faster for a recurring report than a chat question.

Faros AI and Sleuth cover overlapping engineering-intelligence ground, each with its own connector.

DORA inputs are also reachable from the source systems. GitHub and GitLab both expose pull request and pipeline data through documented MCP servers.

Ad hoc questions across teams and periods, answered without building a dashboard view first.

Connect it if you already use LinearB and want ad hoc metric questions answered in chat. Keep official reporting in the dashboard.

LinearB MCP server setup

Every client points at https://mcp.linearb.io/mcp. OAuth sends you to LinearB to sign in and approve access, and needs no key.

For an API key, create one under Settings, API Tokens in LinearB and send it as the x-api-key header.

claude mcp add --transport http linearb https://mcp.linearb.io/mcp

# with an API key instead of OAuth
claude mcp add --transport http linearb https://mcp.linearb.io/mcp \
  --header "x-api-key: <your-linearb-api-key>"

{"mcpServers":{"linearb":{"url":"https://mcp.linearb.io/mcp"}}}

codex mcp add linearb --url https://mcp.linearb.io/mcp
codex mcp login linearb

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"servers":{"linearb":{"type":"http","url":"https://mcp.linearb.io/mcp"}}}

{"mcpServers":{"linearb":{"serverUrl":"https://mcp.linearb.io/mcp"}}}

{"mcpServers":{"linearb":{"command":"npx",
  "args":["mcp-remote","https://mcp.linearb.io/mcp"]}}}

{"context_servers":{"linearb":{"command":"npx",
  "args":["-y","mcp-remote","https://mcp.linearb.io/mcp"]}}}

Read-only access, permissions and security

• The server inherits broad access

  LinearB connects to Git providers, issue trackers and CI, so a connected agent sees a wide slice of engineering activity.

• Engineering metrics describe people

  Cycle time and review latency aggregate individual behaviour, which makes access control a people question as well as a data one.

• API keys are organization-scoped

  A key gives the agent the organization's view rather than one user's, so prefer OAuth for people and keep keys for service accounts.

• No write operations are documented

  The help centre describes querying and reporting only; confirm before granting an agent any role above Viewer.

Troubleshooting

The client cannot find the LinearB server

Check the URL is exactly https://mcp.linearb.io/mcp, the server is enabled in the client, and the client was restarted after the config change.

Authentication keeps failing

Confirm you are signing in to the right LinearB organization and have Company Admin, Editor or Viewer access. For an API key, check it has not been rotated or revoked.

The browser never opens for OAuth

The client may not support OAuth. Use the x-api-key header, or the mcp-remote bridge that LinearB documents for Cline and Zed.

A figure does not match the dashboard

Ask one question at a time with a clear timeframe, team and repository, then compare with LinearB directly. LinearB support asks for the saved prompt and response.

LinearB MCP server: Reference

What engineers report

Public engineering discussion of the server is thin. LinearB's help centre is the main written record, and its timeline is below.

Help-centre articles and the MCP category retrieved 3 October 2026, with article dates from their structured metadata. The March 2026 tool list comes from the Internet Archive capture of 7 March 2026. GitHub searched for linearb and mcp on 3 October 2026.

Should you connect an agent to LinearB?

Yes. Best for engineering managers preparing retrospectives, release reviews and team comparisons.

Strongest fit

Yes, if your organization already runs LinearB and managers want to ask metric questions in an AI client.

Main advantage

It is hosted with OAuth and documented for eight clients, so setup is one URL and a browser approval.

Main weakness

Current docs list data domains rather than tool names, so you learn the exact tools only after connecting.

Operational risk

People metrics reach whoever holds an organization API key, and summaries can drift from the dashboard.

Frequently asked questions