PlugWiki
Currents logo

Currents MCP Server: Setup for Claude, Cursor and Other Apps

OfficialLocal
Gethttps://mcp.currents.dev/mcpcurrents-dev/currents-mcp

Lets Claude and Cursor read your Currents CI test runs, find flaky tests and manage quarantine rules and webhooks.

This is Currents' own server for its test dashboard. Your assistant can look up failed runs, see why a test broke, check how flaky a test has been, and file Jira issues from a failure. It runs on your computer with a Currents API key, and the same server is also available as a hosted address. It can also change things in Currents: it can delete runs, rules and webhooks, so check what the assistant plans to do before you approve it.

Good for

  • Developers who want Claude or Cursor to fix tests that failed in CI
  • QA engineers tracking flaky tests and slow spec files
  • Teams that manage quarantine rules and webhooks in Currents
  • Anyone who files Jira bugs from failed test runs

Works in

12 should work

Tap an app to see how to set it up and any known problems.

Install Currents MCP

Claude Code

  1. 1

    Copy this prompt.

    Add the Currents MCP server to Claude Code. Run this command:
    
    claude mcp add currents -e CURRENTS_API_KEY=<CURRENTS_API_KEY> -- npx -y @currents/mcp
    
    Ask me for CURRENTS_API_KEY and only put the values in that config, nowhere else.
    
    Then check that it loaded and tell me if I need to sign in or restart.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

Built from the package data we hold. Not yet tested by us.

Tools

  • Shows why a CI run, spec file or test failed, with errors and steps

    Returns the errors, steps and files of the failed tests in a run, a spec file or a single test. It is the same content as the Fix button in the Currents dashboard, and the server tells the assistant to call it first when tests fail.

    Try: “Why did the checkout tests fail in the latest CI run on main?”

    Network · currents-get-context

  • Lists recent test runs for a project, with filters

    Lists runs for a project and filters them by branch, tags, status, date range, author or free text such as a commit message or pull request title. It needs a project ID, which the assistant can look up first.

    Try: “Show the failed runs on the main branch from the last 3 days.”

    Network · currents-get-runs

  • Shows one run: status, commit, branch and spec files

    Returns a run's status, commit, branch, groups and spec files, each with its instance ID, and how many attached files the run has.

    Try: “Give me the details of this run: https://app.currents.dev/run/abc123”

    Network · currents-get-run-details

  • Finds a run from details such as a commit or build

    Looks up a run by query parameters instead of its ID, for when you know the branch, commit or CI build but not the run.

    Try: “Find the run for commit 4f2a9c1 in the web-app project.”

    Network · currents-find-run

  • Shows the raw result of one spec file execution

    Returns each test's state and attempts for one spec file execution, plus links to its artifacts such as screenshots.

    Try: “Show every attempt of the tests in checkout.spec.ts for this run.”

    Network · currents-get-spec-instance

  • Shows the history of one test across runs

    Returns past executions of a single test, found by its signature. You can filter by date, branch, tags, author, status or flaky state.

    Try: “Has the 'applies coupon' test failed on main in the last two weeks?”

    Network · currents-get-test-results

  • Works out the ID used to look up one test

    Builds the unique signature for a test from the project, spec file path and test title. The history tools need this signature.

    Try: “Get the signature for the 'logs in' test in auth.spec.ts.”

    Network · currents-get-tests-signatures

  • Ranks tests by failures, flakiness or duration

    Returns aggregated metrics per test for a date range. You can sort by failures, flakiness, duration or runs, and filter by spec, title, tags, branches or authors.

    Try: “Which 10 tests were flakiest this month?”

    Network · currents-get-tests-performance

  • Ranks spec files by failure rate, flake rate or duration

    Returns metrics per spec file for a date range, sortable by average duration, failure rate, flake rate or timeout rate.

    Try: “Which spec files are the slowest on average this quarter?”

    Network · currents-get-spec-files-performance

  • Groups errors across a project

    Returns aggregated error metrics for a project over a date range, so you can see which errors repeat most.

    Try: “What are the most common errors in our CI tests this week?”

    Network · currents-get-errors-explorer

  • Shows project-wide run and test metrics

    Returns summary numbers for a project over a date range, such as how many runs passed or failed.

    Try: “Give me a pass/fail summary for the web-app project over the last 30 days.”

    Network · currents-get-project-insights

  • Lists pull requests with their test runs

    Returns one card per pull request, with the runs grouped under it.

    Try: “Which open pull requests have failing test runs?”

    Network · currents-list-pull-requests

Read from the source code

Permissions

high risk · 4 things to watch

High risk

The assistant can read your test results and also delete runs, rules and webhooks, change quarantine rules and make public links to test results. Use a key from a project you can afford to change, and approve each write or delete yourself.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It can delete runs, rules and webhooks

    Delete tools remove a run and all its data, archive a rule or delete a webhook. Ask the assistant to confirm before it calls them. github.com

  • It can change how tests run

    Creating, updating, disabling or enabling a rule can quarantine or skip tests in later CI runs. Read the rule before you approve it. github.com

  • It can make public links to test results

    Share and evidence links open without signing in until they expire. Create one only when you mean to hand results to someone. github.com

  • Your test data goes to your AI app

    Results, errors and CI details pass through whichever assistant you connect. The README tells you to vet any AI service you give access. github.com

Security advisories

No published security advisories as of Nov 1, 2026.

Technical details
Turning on read-onlyNo read-only mode is documented. The local server turns on every tool whatever the key allows, because it cannot see its key's scope.
Talks toCurrents API (api.currents.dev)
Sign-in and keysYour Currents API key goes in the CURRENTS_API_KEY setting of your MCP config. The hosted address takes the same key as a Bearer token.
Scopes
  • Currents API key (CURRENTS_API_KEY)

Questions

Is the Currents MCP server official?

Yes. It is published by Currents (the currents-dev organization) for its own test dashboard.

Do I need a Currents API key?

Yes. Create one by following the Currents API key guide, then put it in the CURRENTS_API_KEY setting shown in the install box.

What is the Currents MCP server URL?

The hosted address is https://mcp.currents.dev/mcp. It needs your Currents API key sent as an Authorization Bearer header. VS Code's install link instead opens a Currents sign-in page, so no key is needed there.

Can it be read-only?

No read-only switch is documented, and the local server turns on all tools. Use a key for a project where changes are acceptable.

Can it delete my test runs?

Yes. It has tools that delete a run, archive a rule and delete a webhook, so approve those calls one at a time.

Does it work with Claude Code and Cursor?

Yes. The README gives setup for Cursor, Claude Code, Claude Desktop, VS Code with GitHub Copilot and Copilot CLI. Follow the tab for your app in the install box.

Does it include skills?

Yes, but separately. The repository also ships three skills for fixing failing tests and collecting evidence, and the server lists them as resources. The skills are not required for the tools to work.

Information

updated Nov 2, 2026

Publisher
Official
Package
@currents/mcp
Registry
npm
Endpoint
https://mcp.currents.dev/mcp
Transport
stdio, Streamable HTTP
Auth
API key
Version
1.0.0
License
Apache-2.0
Last commit
Oct 30, 2026
Open issues
0
Listed in
Official MCP registry

WebsiteGitHubnpmOfficial MCP RegistryChangelogIssues

Sources

Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.

Show all 2
  1. github.com · github.com
  2. github.com · github.com

Facts from the official MCP registry, GitHub, npm, PyPI and the Docker catalog. Updated Nov 2, 2026.