kadeep mcp
OfficialAllows running KaDeep Studios AI end-to-end tests on Android mobile platforms as part of release-readiness checks.
Allows running KaDeep Studios test cases on the Firefox browser, including specifying Firefox as the browser for specific tests.
Allows running KaDeep Studios AI end-to-end tests on iOS mobile platforms as part of release-readiness checks.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@kadeep mcprun the smoke suite and tell me which tests fail and why"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
KaDeep Studios is an autonomous release-readiness QA agent. You give it your product and the user journeys that make revenue, and it runs them on web, mobile and desktop. What comes back is a go or no-go, not a test report.
KaDeep CLI brings KaDeep Studios to the command line, as two open-source npm packages:
Package | Command | What it does |
| The KaDeep Studios command line and MCP server. Type | |
| KaDeep Release Gate, a go / no-go quality gate for CI/CD. It reads |
Contents
Related MCP server: squish-mcp
Quick start
You need Node.js 20+ and a KaDeep Studios account (get started for free).
npx kadeep # the interactive home: sign in, pick a project, then ask the agent or type /
npx kadeep run --suite smoke # or one command at a time: run a suite and wait for the verdict
npx kadeep init # add a releasegate policy + GitHub Actions workflow to this repoAfter kadeep init, every pull request runs npx releasegate and gets a go / no-go:
KaDeep Release Gate 0.3.0 · Engineering release confidence.
project p_9a118f0499704752be72 · commit 8c3f2d1 on feature/checkout PR #214 · mode enforce
▸ Smoke
✓ Smoke: 12/12 passed (2m 41s)
▸ Checkout
✗ Checkout: 3/4 passed (1m 05s)
✗ Pay by card [DEFECT] · Expected "Order placed" to be visible
NO-GO · 1/2 required checks passed · 3m 47sWhat it looks like
A terminal UI built on the dot, like the KaDeep Studios logo: every test is a dot, and the release verdict is drawn in dots. It has no dependencies, and output stays plain in CI.
Features
You want to… | Command |
Drive everything from one place in the terminal, like Claude Code |
|
Ask the KaDeep agent why something failed, or have it act | Type the question in the home, or |
Run AI end-to-end tests from the terminal |
|
Run specific test cases on another browser or screen size |
|
See why a test failed, step by step |
|
Browse projects, suites, test cases and defects |
|
Block a release when required checks fail |
|
Try the gate first without blocking anyone |
|
Set up a repository (policy, CI workflow, agent config) in one step |
|
Let Cursor or Claude Code run and read your tests |
|
Check localization quality in CI |
|
Script anything, parse the output |
|
Publish test results to CI dashboards |
|
Watch runs live in the terminal, plain in CI | Rich view in a terminal; |
From commit to go / no-go
flowchart LR
A["Pull request<br/>or push"] --> B["CI runs<br/>npx releasegate"]
B --> C["Reads<br/>releasegate.yml"]
C --> D["KaDeep Studios runs the checks<br/>suites · test cases · localization"]
D --> E{"Verdict"}
E -->|GO| F["Merge and ship"]
E -->|NO-GO| G["Build fails<br/>report + PR annotations"]
E -.->|shadow mode| H["Reported only<br/>never blocks"]GO: every required check passed.
NO-GO: a required check failed or errored. An empty suite is never a pass.
Shadow mode: the verdict is reported but never fails the build, so you can watch the gate for a while before letting it block.
The report lands in .releasegate/ (report.json, report.md, junit.xml). On GitHub it also goes into the job summary and annotates failing checks.
How it works
flowchart TB
dev["Developer terminal"] --> kadeep
agent["AI coding agents<br/>Cursor · Claude Code"] -->|MCP over stdio| kadeep
ci["CI/CD pipeline<br/>GitHub Actions · GitLab CI · CircleCI"] --> rg
rg["releasegate<br/>policy · verdict · reports"] -->|npm dependency| kadeep
kadeep["kadeep<br/>CLI · MCP server · client library"] -->|HTTPS| api["KaDeep Studios<br/>api.kadeep.ai"]
api --> surfaces["Web · iOS · Android · Desktop"]kadeepis the foundation. It contains the CLI, the MCP server and the client library: sign-in, runs, localization and CI detection. It has no runtime dependencies.releasegateis built onkadeep. It adds the policy file, the go / no-go decision, shadow mode and the reports. It depends onkadeep, never the reverse.Only HTTPS calls to the KaDeep Studios API. Tests run in KaDeep Studios, on the model key your team configured there. The CLI itself never calls an AI model and collects no telemetry.
Runs are jobs on KaDeep's queue. The CLI starts them and polls, so no request stays open while a suite runs. A long suite survives proxies with idle timeouts, and a CI network blip costs one retried poll:
sequenceDiagram
autonumber
participant CI as releasegate (CI job)
participant API as KaDeep Studios API
participant Q as KaDeep workers
CI->>API: POST /api/ci/:project/run { async: true }
API-->>CI: 202 · job id
API->>Q: queue the suite
loop every few seconds
CI->>API: GET /api/ci/:project/jobs/:id
API-->>CI: running · 3/12
end
Q-->>API: suite run finished
CI->>API: GET /api/ci/:project/jobs/:id
API-->>CI: done · results per test
Note over CI: verdict → report.json · report.md · junit.xml → exit codeUse cases
Run end-to-end tests from the terminal
kadeep use "Acme web" # default project
kadeep run --suite smoke # ✓ / ✗ per test, exit 1 if any fail
kadeep runs show r_bf6382d90aab # every step, the verdict and why
kadeep run --suite regression --no-wait # queue it, come back later
kadeep jobs show job_8ad6a6bb92 --waitAdd a release gate to GitHub Actions
kadeep init generates .github/workflows/releasegate.yml. Its core:
name: KaDeep Release Gate
on:
pull_request:
push:
branches: [main]
jobs:
releasegate:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: 22
- run: npx -y releasegate@^0.1
env:
KADEEP_CI_TOKEN: ${{ secrets.KADEEP_CI_TOKEN }}The policy file sits next to it:
# releasegate.yml
version: 1
project: p_9a118f0499704752be72
mode: shadow # switch to enforce once you trust the verdicts
checks:
- suite: smoke
- tests: [checkout-works]
viewport: mobile
- localization:
locales: [hi-IN, ar-AE]GitLab CI, CircleCI, Bitbucket Pipelines, Buildkite, Azure Pipelines and Jenkins work the same way. See releasegate: CI setup.
Let AI coding agents run your QA (MCP)
claude mcp add kadeep -- npx -y kadeep mcp # Claude Code{ "mcpServers": { "kadeep": { "command": "npx", "args": ["-y", "kadeep", "mcp"] } } }That second snippet goes in Cursor's .cursor/mcp.json. Your agent can then answer "which smoke tests fail and why?" or "run the checkout tests" by itself. It gets 14 tools: projects, suites, test cases, runs, jobs, defects and the localization gate.
Gate localization in CI
kadeep loc push --file locales/en.json --translate # new source strings
kadeep loc validate --locale hi-IN --locale ar-AE # exit 1 with the reasons
kadeep loc pull --asset en.json --locale hi-IN # approved translations onlyWhich package do I need?
If you… | Use |
work in a terminal or write scripts |
|
want Cursor, Claude Code or another MCP client to use KaDeep |
|
want CI to pass or fail a build on KaDeep results, with a policy and reports |
|
just want to run one suite in CI and fail on red |
|
Terminology
Term | Meaning |
KaDeep Studios | The QA platform and app where your projects, test cases and runs live |
KaDeep CLI | This repository's command line, the |
KaDeep Release Gate | The CI gate, the |
Test case | One user journey KaDeep runs, e.g. checkout works. |
Suite | A named group of test cases, e.g. smoke |
Run / suite run | One execution of a test case / of a suite, with a verdict and evidence |
Job | A queued or running execution. The CLI follows jobs until they finish |
CI token | A project's token for pipelines ( |
GO / NO-GO | The gate's verdict. Shadow mode reports it; enforce mode fails the build on NO-GO |
FAQ
A release gate is an automated check in your CI/CD pipeline that decides whether a change may ship. releasegate makes that decision from KaDeep Studios results for the exact commit under test: suites, test cases and localization. It returns GO or NO-GO, with a report of why.
Yes. kadeep and releasegate are open source under Apache-2.0. They connect to KaDeep Studios, where your tests run, and you can get started for free.
Any pipeline that can run Node.js 20. The commit, branch, PR number and run link are read automatically on GitHub Actions, GitLab CI, CircleCI, Bitbucket Pipelines, Buildkite, Azure Pipelines and Jenkins. Anywhere else, the gate falls back to git.
Yes. kadeep mcp is a Model Context Protocol server over stdio. Any MCP client can list your tests, run them, wait for the verdict and read failures step by step.
It only makes HTTPS calls to the KaDeep API you point it at (https://api.kadeep.ai by default), with no telemetry. Your session is stored in ~/.config/kadeep/config.json, readable only by you. In CI, a per-project CI token is used instead of a personal login.
Yes. Set mode: shadow in releasegate.yml (kadeep init starts there). The gate runs every check, writes the report and says "would have blocked", but always exits 0. Switch to mode: enforce when its verdicts match your judgement.
Development
git clone https://github.com/kadeep-ai/kadeep-cli.git && cd kadeep-cli
npm install
npm run ci # typecheck (JSDoc + checkJs), tests on the local Node, tarball checkpackages/
kadeep/ CLI, MCP server and client library (no runtime dependencies)
releasegate/ KaDeep Release Gate (depends on kadeep and yaml)
scripts/
check-packs.mjs guards what npm publishesSources: plain ES modules with JSDoc types. There is no build step, so what is in
packages/*/srcis what ships.CI: runs the tests on Node.js 20, 22 and 24.
Tarball check:
scripts/check-packs.mjsfails a publish if anything other thanbin/,src/,README.mdandLICENSEwould ship, or if a file contains a secret or a source map.Releases: both packages share one version. Publish
kadeepfirst, thenreleasegate.
Found a bug or have an idea? Open an issue.
License
Apache-2.0 © Kadeep Technologies. KaDeep Studios is a product of Kadeep Technologies (kadeep.ai).
This server cannot be deployed
Maintenance
Related MCP Connectors
Direct access to Cypress tests results and accessibility reports in your AI workflow.
Code intelligence platform for AI agents. 20 tools for architecture, security & impact analysis.
Run, debug, and triage tests from your IDE using natural language, no dashboard switching, no manual data transfers. The TestMu AI (formerly LambdaTest) MCP Server is a single remote server exposing four tool suites: HyperExecute — analyze your project, generate YAML configs and test runner commands, then monitor jobs and sessions. Automation — pull a TestID's details plus command, network, and console logs into one chat for instant root-cause analysis. Includes mobile app upload. SmartUI — explain pixel, layout, DOM, and perceptual changes in a visual regression run, with context-aware React/HTML/CSS fixes. Accessibility — audit any public URL or a local React app against WCAG and get ready-to-apply remediation steps. Connects over https://mcp.lambdatest.com/mcp using OAuth 2.1 — no API keys in your config. One-click install in Cursor; works with Claude, GitHub Copilot, Cline, and any MCP client. Tests execute on the TestMu AI cloud: 3,000+ browsers and 10,000+ real devices.
Develop, manage, and debug Railway projects, services, and deployments from within agents.
Related MCP Servers
- AlicenseNot gradedqualityAmaintenanceOpen-source terminal AI agent with 600+ tools, 35 specialist agents, and 20 providers. Plug kbot into Claude Code, Cursor, VS Code, Zed, or Neovim as an MCP tool provider.11 npm16MIT

squish-mcpofficial
FlicenseAqualityFmaintenanceEnables AI agents to run and create Squish test scripts and test suites and analyze the results.1913-- AlicenseNot gradedqualityAmaintenanceAutonomous spec-to-product coding-agent CLI. Its MCP server exposes 34 tools over stdio: project state and task-queue ops, memory retrieve/store, code search, quality and verification reports, repo hotspots/co-changes, and structured findings/learnings.7,848 npm1,077Business Source 1.1
- AlicenseNot gradedqualityDmaintenanceEnables AI models (Claude, ChatGPT, GitHub Copilot) to run and analyze local tests, rerun failures, and orchestrate QA workflows using existing UI and API test frameworks.8 npmMIT