Overview

Marker.io MCP lets AI agents access your issue reports, read technical context, view screenshots, and take actions like updating status or adding comments, directly from any MCP-compatible tool.

💡 Beta

MCP integration is currently in beta and free to use for all teams. Pricing will be announced when the feature exits beta.

Setting up

For step-by-step instructions for ChatGPT, Claude, Cursor, VS Code, Windsurf, JetBrains, and more, see MCP setup guides.

The MCP server URL is:

https://connect.marker.io/mcp

There are two ways to authenticate:

OAuth - most clients handle this automatically. On first use, your AI tool opens a browser window where you approve access to your Marker.io account.
Access token - for clients that don't support OAuth, generate a bearer token from your Marker.io account settings.

Quick start with OAuth (e.g. Claude Code)

claude mcp add --transport http marker-io https://connect.marker.io/mcp

Quick start with access token (e.g. GitHub Copilot)

{  "servers": {    "marker-io": {      "type": "http",      "url": "https://connect.marker.io/mcp",      "headers": {        "Authorization": "Bearer YOUR_TOKEN"      }    }  }}

View all setup guides.

We also have CMS-specific guides for Shopify, Webflow, Contentful, Adobe Experience Manager, Prismic, and Salesforce Commerce Cloud.

Available tools

24 tools. OAuth 2.1 authentication. Works with any MCP-compatible client.

What the agent can do:

Read projects, issues, screenshots, attachments, console logs, network requests
Act on issues: update status, priority, type, assignee
Collaborate with the team: post comments, @mention teammates
Monitor website accessibility: scores, failing checks, and the elements behind them

Projects

Tool	Description
`projects_list`	List Marker.io projects for the current team. Optional filters: `searchText`, `archived`, `limit`.
`project_get`	Get details of a specific project by `projectId`.
`project_list_users`	List users (members) of a project by `projectId`. Returns id, name, email, role, avatar.

Issues

Tool	Description
`issues_list`	List issues for a project. Filter by `status`, `priority`, `searchText`. Supports pagination (`limit`, `offset`).
`issue_get`	Get full details of a specific issue by `issueId`.
`issue_update_status`	Update an issue's status (`open`, `in-progress`, `waiting-for-customer`, `resolved`, `archived`).
`issue_update_assignee`	Assign or unassign an issue. Use `project_list_users` to find valid user IDs.
`issue_update_priority`	Update an issue's priority (`none`, `low`, `medium`, `high`, `blocker`).
`issue_update_type`	Update an issue's type. Use `list_issue_types` to find valid type IDs.
`list_issue_types`	List available issue types for a project.
`comment_create`	Add a comment to an issue. Supports inline `@userId` mentions and `Everyone` or `Member-only` visibility.

Issue context and media

Tool	Description
`issue_get_context`	Get technical context: browser, OS, screen size, console log summaries, network request summaries.
`issue_get_console_log`	Get full details of a specific console log entry (by `logIndex` from `issue_get_context`).
`issue_get_network_request`	Get full details of a specific network request (by `requestIndex` from `issue_get_context`).
`issue_get_screenshot`	Get a signed URL for the issue's screenshot.
`issue_get_attachments`	List issue attachments with signed download URLs.

Website monitoring

Each project monitors one website for accessibility. These tools read the score, the failing checks, and the elements behind them.

Tool	Description
`monitor_summary`	Get a project's accessibility overview: current score, recent change, and pass, fail, and ignored check counts. Start here.
`monitor_pages_list`	List monitored pages with each page's score, failing-check count, and last scan status. Worst pages first.
`monitor_checks_list`	List accessibility checks (WCAG rules), defaulting to failing checks ranked by score impact.
`monitor_check_guide`	Get the step-by-step guide for fixing a specific check, by `checkId`.
`monitor_elements_list`	List failing elements, grouped by fingerprint, with occurrence and page counts.
`monitor_element_get`	Get one element's full HTML, suggested fix, failing check, and the pages it appears on.
`monitor_element_screenshot`	Get a rendered screenshot of a failing element.
`monitor_scans_list`	List scan history with scores over time to see the accessibility trend.

Data and privacy

MCP uses OAuth 2.1 with PKCE for authentication. Your credentials are never shared with the AI tool.

Access tokens are scoped to mcp:tools and expire automatically
Tokens can be revoked at any time from your Marker.io account
All communication is encrypted via HTTPS
The AI tool only accesses projects and issues your account has permission to view

Learn more in AI data security & privacy.