Discover the Best AI Tools

pytest-jev adds probability-based semantic assertions to pytest and batches claims about one output into a Jev request. It requires Python 3.10+, pytest 7.4+ and an API key; without a key the affected tests are skipped, and model judgments are not deterministic truth.

Ground Zero evaluates response grounding, correctness against expected answers and instruction following using Jev judgments. It is an alpha library requiring OpenRouter access; schemas may change and inconsistent judge outputs need manual validation.

replicant-mcp lets coding assistants build Android packages, operate emulators, inspect logs and interact with app interfaces. It requires Node.js 18+ and the Android development toolchain; device-changing actions include uninstalling apps and clearing data.

Tactual models screen-reader navigation cost from browser snapshots and keyboard probes, with CLI, CI and MCP access. It requires Node.js 20+; simulated assistive-technology profiles complement, rather than replace, real screen-reader and manual validation.

Repo Test Architect audits repository structure, existing tests and framework signals to rank testing work and produce plans through CLI or MCP. Adapter coverage is bounded; native test generation remains deferred rather than implemented.

Python evaluation and guardrail library batching typed questions about agent traces, tool use, grounding and safety. Supports configured hosted Jev providers, local decision models or chat-model fallbacks.

MockHero generates synthetic relational test data from schemas, natural-language requests or templates, exporting JSON, CSV or SQL. Its hosted API requires a key; generated records are test fixtures, not verified real-world data.

mobile-device-mcp exposes screenshots, touch actions, UI inspection and optional visual analysis to coding assistants. It requires Node.js 18+ and mobile tooling; real iOS devices are still planned, visual AI needs provider access, and the code uses BSL 1.1.

MCP server for running web, mobile and API test suites, inspecting failure artifacts and producing follow-up test plans. Runner-specific tools such as Maestro, Newman or browser binaries must be installed separately.

MCP layout inspector measuring rendered DOM and text rectangles to flag overlap, overflow and leftover placeholders without baseline images. Inspection only; currently requires building from source and a compatible Chromium browser.

Uses Playwright to measure rendered UI spacing and element sizes, returning MCP or HTML reports. Does not edit source or fail CI; non-Tailwind pages receive only the supported accessibility checks.

Captures real agent verification runs as evidence bundles and checks them with a deterministic CI gate. The gate uses no AI or network; independent model attestation is optional.

Deterministic .NET test-solution indexer with a SQLite map and MCP queries for steps, scenarios and dependencies. Requires .NET SDK 8+ and indexing your solution; it supplies code context rather than running an AI model or proving tests pass.

MCP server and CLI for code-aware Playwright browser tests, regression history and screenshot reports. Requires Node.js 20+ and a browser installation; checks are heuristic and the client agent supplies test reasoning, rather than an internal LLM.

MCP for disposable inboxes and extracting verification codes or magic links through ZeroDrop. The free sandbox needs no API key and deletes inboxes after 30 minutes; workspace access uses a key. Email processing occurs on the external Cloudflare-backed service, not solely on the local MCP client.

MCP combining disposable test email and rented UK phone numbers to check signup and verification flows. Email creation is public but rate-limited and mailboxes last up to 60 minutes. Phone tools require an sms-florin API key and balance; numbers and received SMS are real, not a simulated telecom service.

MCP connector for renting UK SMS numbers and checking received messages during authorized signup or verification tests. Requires a sms-florin API key and funded balance; renting debits that balance. Number availability and service compatibility depend on the provider, and this connector uses real numbers rather than mocked test messages.

MCP testing companion for Cline that sets up project rules, plans scenarios, selects templates and classifies failures using a project memory bank. The host agent writes and runs tests; this is not an independent test runner. Automation and approval settings are opt-in, and setup writes project configuration.

Experimental local MCP/CLI browser tool returning semantic page digests, typed action diffs and assertions over retained evidence. Requires Node.js 22+ and its own Chromium session. It cannot reuse normal browser logins; canvas, frames and closed shadow roots have blind spots, and no observable change does not prove task success.

MCP server exposing Selenium browser actions and assertions, recording sessions to generate pytest, TestNG, JUnit or Cucumber test code. Requires Python 3.10+ and Chrome; Selenium Manager handles the driver. Generated tests still need review and execution in the target project; code generation alone is not successful test validation.

Python browser automation and testing framework with pytest integration, recording, assertions, reports and MCP access. Supports browser interaction and scraping through Selenium/CDP modes. Requires a compatible browser environment; stealth and CAPTCHA-related capabilities depend on site conditions and do not guarantee universal access or correct tests.

MCP connector to Prufa’s hosted website auditing and QA service, returning graded findings, evidence and fix hints and supporting flows and regression monitoring. API credentials and service allowances apply; advanced tools are plan-gated. The open-source connector is not the hosted audit engine, and automated grades do not certify legal compliance or complete defect coverage.

Playwright-based MCP server for website QA, persistent login sessions, assertions, network mocking, visual comparisons and accessibility/SEO audits. Requires Python 3.11+ and Chromium. Interactive login lets a person complete 2FA or CAPTCHA; automated findings and Lighthouse checks do not establish complete accessibility or functional correctness.

An alpha suite for self-contained HTML test replays and local browser-session inspection/control via MCP. Tracelane supports WebdriverIO and Playwright; Cypress remains planned. Peek requires Node.js 22+ and consent for live control. Local recordings may still be shared by the connected AI client.