accessiBe Help Center

How to connect AI clients to accessFlow with MCP

  • Updated

Connect AI clients to accessFlow with an MCP server to find urgent accessibility issues, get WCAG-based guidance, and streamline remediation.

Overview

The accessFlow Model Context Protocol (MCP) server provides intelligent accessibility issue analysis and remediation guidance. It connects AI clients to the accessFlow platform to help developers identify, prioritize, and fix accessibility issues in their web applications.

The accessFlow MCP Server enables AI clients (like Copilot) to:

  • Identify priority issues: Get the most urgent accessibility issues, ordered by severity and impact.
  • Provide fix guidance: Generate detailed remediation instructions with code examples.
  • Align with WCAG: Reference specific WCAG guidelines and criteria for each issue.
  • Streamline workflow: Get detailed accessibility insights and fix recommendations directly in your IDE.

API key setup

Generate a token in accessFlow to to set the API key.

To generate a token

  1. Go to your Profile menu in top right corner and select Token management
  2. Select Generate token.
  3. Enter a token name.
  4. Select AI MCP token and then Generate token.
  5. Copy the token and store it securely. You won't be able to see it again.

Configuration

Add the configuration to your MCP client (this varies by client), usually in the file mcp.json.

Configure the accessFlow MCP server locally

Enter your copied token into API_KEY:

{
  "mcpServers": {
    "flow-mcp": {
      "command": "npx",
      "args": ["-y", "accessflow-mcp-server"],
      "env": {
        "API_KEY": "your-accessflow-api-key",
        "DOMAIN": "<https://your-app-domain.com>"
      },
      "type": "stdio"
    }
  }
}

Configuration parameters

ParameterRequiredDescription
API_KEYYesYour accessFlow API key for authentication
DOMAINYesThe domain of your application being analyzed (e.g., https://example.com)

Connect to the accessFlow MCP server remotely

Connect remotely to the accessFlow MCP server: flow-mcp.accessibe.com

"flow-mcp": {
      "type": "http",
      "url": "<https://flow-mcp.accessibe.com/mcp>",
      "headers": {
        "Authorization": "Bearer <FLOW_MCP_API_KEY>",
        "X-Domain": "mysite.com",
      }

Configuration parameters

ParameterRequiredDescription
AuthorizationYesYour accessFlow bearer token and API key for authentication
X-DOMAINYesThe domain of your application being analyzed (e.g., https://example.com

MCP server commands

Once configured, you can run commands through your MCP client to identify accessibility issues and get remediation guidance. You can use natural language or the command name. 

Get most urgent issues

getMostUrgentIssues

Retrieves the most critical accessibility issues, prioritized by:

  • Severity: Extreme → High → Medium → Low
  • Site impact: Number of locations affected across the site
  • Page impact: Number of occurrences on individual pages

Parameters 

None (uses the configured domain)

Returns

JSON data with structured issue information including severity, WCAG level, occurrence counts, and unique identifiers.

Example

"Show me the most urgent accessibility issues for my site"

Get issue remediation guidance

getIssueRemediation

Provides detailed remediation guidance for a specific accessibility issue.

Parameters

ParameterTypeRequiredDescription
issueDisplayNamestringYesThe unique identifier for the accessibility issue.

Returns

Comprehensive fix guidance including:

  • Problem summary and WCAG references
  • Current problematic code
  • Suggested code fixes with examples
  • Step-by-step remediation instructions
  • Links to tutorials and additional resources

Example

"Get remediation guidance for issue: Decorative-Content-6d277a13ba"

Mark an issue as resolved

resolveIssue

Marks the specified issue as resolved in accessFlow.

Note: If the next audit still detects this issue, it will be reopened automatically.

Parameters

ParameterTypeRequiredDescription
issueDisplayNamestringYesThe unique identifier for the accessibility issue.

Example

"Resolve issue id Footer-Landmark-e3c792e8b8"le

Supported clients

This MCP server works with AI agents in IDEs and other MCP-compatible clients:

  • VS Code Copilot: Integrates with Microsoft Copilot in VS Code
  • Cursor: Works with Cursor's AI assistant
  • Other MCP Clients: Any application supporting the Model Context Protocol

Was this article helpful?

0 out of 0 found this helpful