MCP Server

Connect AI Agents to Currents

What is MCP

MCP stands for Model Context Protocol. It's an open pattern, introduced by Anthropic, that provides a consistent way for systems to expose tools and resources that can be used by AI models.

Currents MCP server is a context layer for AI tools that leverage information about Playwright test results, such as failed tests, errors, and more.

Get started

Install our NPM package

npm install @currents/mcp

Setup the MCP Server

Cursor

1. Go to Cursor Settings > MCP > Enable

2. Add the following to your mcp.json

{
  "mcpServers": {
    "currents": {
      "command": "npx",
      "args": [
        "-y",
        "@currents/mcp"
      ],
      "env": {
        "CURRENTS_API_KEY": "your-api-key"
      }
    }
  }
}

Claude

Example Prompt

@folder Tests are failing in CI. Get all the details from the run <runId> fix the failures

Get the runId from the run's "Advanced" tab in the dashboard. Soon, the MCP server will be able to fetch the latest runs for an organization, removing the need for users to provide a specific run id.

Use Cases & Capabilities

Currents MCP server exposes a variety of tools.

Tool	Description
currents-get-projects	Retrieves a list of all projects available.
currents-get-runs	Retrieves a list of the latest runs for a specific project.
currents-get-run-details	Retrieves details of a specific test run.
currents-get-spec-instances	Retrieves execution results of a spec file.
currents-get-spec-files-performance	Retrieves spec file historical performance metrics for a specific project.
currents-get-tests-performance	Retrieves test historical performance metrics for a specific project.
currents-get-tests-signatures	Returns Signature (filtered by spec file name and test name). Allows an agent to find test results of a specific test.
currents-get-test-results	Retrieves test results of a test, supports multiple filters.

---

These tools can be used to provide context to the AI agent about all the details of a run, tests executions, and specs, including historical data like error rate, debugging logs, duration, flakiness, and more.

Here are some examples of AI prompts:

• "Please fix this test"

• "Summarize my last test run"

• "What were the top flaky tests in the last 30 days?

• "What were the slowest specs in the last 7 days?"

• "Please fix all my flaky tests"