"Search for HITL (Human-in-the-Loop or other possible acronyms)" matching MCP servers:

Enables coding agents to see the UI they just changed without flooding their context window, returning a structured text summary (console errors, failed requests, overflow, layout shifts, accessibility violations) plus small crops of only the regions that changed since the last capture. All output is fitted into a chosen token budget, with captures stored on disk and evicted automatically.

Enables a coding agent to keep one JSON feature specification per feature, governed by a progressive schema, and to render it as editable Markdown documents for a product owner. Answers and direct Markdown edits are reconciled back into the validated model through stdio MCP tools, applying coherent changes and reporting those that widen scope or contradict the model.

Provides a stable, side-effect-free MCP tool contract that enables AI workflows to be prototyped, demoed, tested, or dry-run without connecting to databases, APIs, files, or other real data sources. It enables swapping in a real data server later without changing the AI-facing interface.

An MCP server that performs independent verification of artifacts against criteria using a different AI model lineage (codex, OpenAI, or Gemini) to catch defects that same-family checks might miss.

A local MCP server that gathers web captures, image crops, video frames, and copy into structured evidence artifacts, letting a vision-capable agent inspect the returned media and produce evidence-backed critique with scored findings, comparisons, and quality gates. It never converts local metadata into visual judgment, so hosts must supply their own vision reasoning.

Enables AI agents to verify their own work on real Godot 4 projects: edit scenes and scripts safely, launch the game, send real keyboard/mouse input, wait on conditions, read live runtime state, capture screenshots and errors, and run declarative scenario play-tests that return per-step pass/fail reports with the observed failing value.

Enables AI agents and MCP clients to drive real Chrome, Edge, Brave, and Firefox browsers through native automation protocols. It exposes tools for navigation, page snapshots, form filling, clicks, screenshots, console/network inspection, and tab management.

Enables MCP clients to open websites, capture screenshots at multiple breakpoints, and run automated design audits for issues like overflow, small tap targets, low contrast, missing alt text, tiny text, and clipped content.

Enables AI agents to inspect UI screen recordings as token-efficient image sheets with changed regions boxed, single-frame flicker flagged, and speech aligned to frames. Exposes watch_video and zoom_video tools so agents can pull dense keyframes across any time range or one sharp full-resolution frame, all processed locally.

Enables AI agents to test Unity scenes and return review-ready receipts via a hosted remote MCP with tools for playmode checks and method invocation.

Exposes 11 agent-native QA tools over stdio so coding agents can discover a project, analyze risk, select the smallest relevant test set, generate and dry-run tests, triage failures, propose tiered heals, analyze flakes, generate quality reports, and evaluate release readiness. Planning and healing tools stay read-only or propose-only, while execution defaults to dry-run, keeping agent actions safe and evidence-backed.

Run and test agentic systems in isolated Docker sandboxes, varying system prompts, models, and task prompts while capturing full behavior traces via MCP tools.

Enables controlled interception, review, and editing of web_search and fetch_page responses through a native desktop console, supporting live retrieval, synthetic or hybrid results, context sessions, and fault injection for AI agent robustness experiments.

Enables MCP clients and agents to query synthetic football match state, key moments, win probability, and player profiles, and to verify claims and on-screen overlay numbers against match state.

Checks the outcome after an agent handoff, when a request, 200 or ack only shows the handoff. Reads one field of one exact public HTTPS JSON record, with a same-site known-good control, for 20-120 s. Returns PROVED, DISPROVED or INCONCLUSIVE with a signed receipt you can verify. 5 free checks, then USD 1 per completed check, by card (Stripe Checkout) or existing Link wallet. INCONCLUSIVE is free.

Zero-sidecar MCP server embedded directly in the Godot 4 editor as a GDScript EditorPlugin — no Node or Python sidecar. Reflection-driven tools over any engine class, validate-before-write GDScript, scene/script/resource authoring, and a runtime play-test loop (play, screenshot, input, assert).

Lets an assistant read recorded production AI agent runs — steps, model and tool calls with arguments, and outcomes — along with the findings of comparison runs that checked a prompt or model change against those recordings, and verify a signed evidence bundle offline with no account or network. It is strictly read-only: no tool starts a replay or comparison, so nothing can spend money or act without a person deciding.

MCP server for remote debugging of Apps in Toss mini-apps, providing CDP relay and test runner capabilities.

Enables users to query, create, update, and manage Digital.ai Agility assets through MCP tools, with secure Windows Credential Manager authentication and configurable write operations.

Enables AI assistants to operate directly inside a running Unity Editor project, creating scenes, generating scripts, simulating input, capturing screenshots, inspecting performance, and automating editor tasks through its built-in tools and resources.

Enables AI agents to create webhook URLs, wait for incoming deliveries, inspect payloads, and replay or send signed webhook events to a local handler.

MCP server that brings Trustabl static reliability scanning into Cursor, letting agents inventory agents/tools, flag risky patterns, score production readiness, and write SARIF reports.

A reference MCP server that enforces identity-aware tool access using role-based permissions (Reader, Analyst, Auditor) and supports multiple identity routes like Entra RBAC, Copilot Studio, and Okta XAA. It provides read-only access to test evidence with explicit denial paths for unauthorized tools.

Connects AI coding assistants to QA engineering standards, regulatory checklists, and a preflight audit engine. Enables skill discovery and installation, license verification, and safe project inspection with human-in-the-loop confirmation.

Enables accessibility auditing of rendered pages with multi-engine evidence, human-in-the-loop review, and fix verification across agent sessions.

Local or self hosted LLM server for Labgrid hardware-in-the-loop device operations

Enables checking npm packages for known security vulnerabilities using the OSV API before installation. Supports both single package checks and bulk vulnerability scanning for multiple packages at once.

A simple server implementing the Model Context Protocol (MCP) that echoes messages back, designed for testing MCP clients.

Bridges AI models with WinDbg to analyze Windows crash dumps and perform remote debugging through natural language queries, enabling execution of debugger commands and automated crash analysis.

Enables coding agents to audit public or localhost pages, gather structured SEO evidence, and retrieve compact reports, finding details, and next actions for triage.