Independent analysis · 2,500 submissions · four equally weighted criteria
WebMCP Rankings
Every WebMCP Challenge submission, read blind and scored on the four official criteria. Read the score, then the reason: open any project to inspect both reviewers' reasoning.
| Project | Aggregate | Creativity |
|---|---|---|
| Physical AI WebMCP Command Center An Agent-Native Control Layer for Autonomous Warehouses. | 38 | 9 |
| Handrail Groceries and pharmacy by voice for people with low vision. The agent does the clicking through WebMCP; the person says yes. | 38 | 9 |
| Alza A floor plan is coordinates, not buttons. Alza publishes its geometry as 31 typed tools, so your agent draws the plan, checks its own work, | 38 | 10 |
| Incident Command A WebMCP safety simulator where a browser agent investigates production incidents, but the dangerous tool to act is never registered until a | 38 | 10 |
| PaperVeil A browser-local medical appeal desk where agents get capabilities—not identity—with one-field consent, adversarial leak blocking, and an aud | 38 | 10 |
| Grenz: a policy layer for WebMCP An AI agent can now act on a site as you, in your browser. Grenz decides what it may do: the site's own rules, approvals a script can't fake | 38 | 10 |
| Mace A clerk's bench for Robert's Rules where the WebMCP tool set IS the motion stack — an action that is out of order does not exist to be calle | 37.5 | 10 |
| Pillbox A caregiver and a pharmacist each get an agent on one medication list, every flag quoting the FDA label. | 37.5 | 9 |
| Substrate A radiology viewer built so an AI agent can do everything except read the scan. | 37.5 | 10 |
| AGENT CONTROL PLANE The website-enforced authority layer for WebMCP agents. | 37 | 9 |
| CareBridge: humans in control of multi-page agent workflows Real work spans many pages, and agents tend to just finish it for you. CareBridge keeps you in control. The agent does the busywork and you | 37 | 9 |
| FlightSweeper A browser agent buys a sandbox flight within exact, revocable limits set by the traveler. | 37 | 9 |
| Floortris WebMCP agents turn dimensioned sketches into rule-checked 2D and 3D rooms, while people compare proposals and control every final change. | 37 | 9 |
| Ingram Sheets A spreadsheet whose engine runs in your browser and publishes twelve tools to whichever agent is driving the tab. Spreadsheets for models, n | 37 | 9 |
| Ops Co-pilot An AI-friendly infra dashboard where agents check real service health and act but every risky action needs a human's explicit approval befor | 37 | 9 |
| Overhead A web-native canvas powered by WebMCP that turns any agent into a cloud architect. Design alongside it live while it calculates real-time in | 37 | 9 |
| Paris ICC Procedure-grounded rail incident decisions, powered by WebMCP and controlled by operators. | 37 | 9 |
| REWIND A transaction layer for AI agents: preview and guard changes before execution, require human approval when needed, and revert committed acti | 37 | 9 |
| RoomCraft Design a room with an agent in seconds. | 37 | 9 |
| ROUGH//CUT An in-browser video editor you and your AI agent can use together | 37 | 9 |
| Sightline The human-approval layer between an agent and a live page. The agent finds what holds a page back, drafts the fix, and stops until you say y | 37 | 9 |
| Sigmora Live Relay Stay live while an agent builds, reviews, and verifies the launch moment with you. | 37 | 9 |
| Verified Mission Control Goal Contracts for the Agent-Native Web | 37 | 9 |
| LAST DOOR When evidence changes, the WebMCP tools change with it. | 37 | 10 |
| Onionskin A floor-plan editor where an AI agent can propose changes but only a human can commit them. Not by policy: a tool that touches the committed | 37 | 10 |
| PARALLAX SPATIAL MCP PARALLAX is a Spatial MCP that lets autonomous agents safely understand, change, re-observe, and verify 3D environments through governed act | 37 | 10 |
| Paper Trail Say it once - one sentence about your life event, and every agency, insurer, and employer form gets filled, cross-checked, approved by you, | 37 | 9 |
| Gridfall One city. One hospital. One chance. GRIDFALL turns WebMCP into a real-time crisis room where a human and an AI agent diagnose cascading fail | 37 | 10 |
| Roadway Design Compiler Agent-native roadway design in the browser, powered by 39 WebMCP tools, a real geometry kernel, terrain-aware intelligence, and seamless ven | 37 | 9 |
| TRACE The canvas is the prompt: a living reasoning surface shared by people and AI agents. | 37 | 10 |
| KnownGate The food verification layer for people's agents: state your family's rules and KnownGate rules products, menus, and plans against read-and-d | 37 | 9 |
| OVERTURN Health insurers deny 1 in 5 claims and almost no one fights back. Overturn pairs you with an agent that builds the appeal through WebMCP too | 37 | 9 |
| Deputy Your agent learns the way !! | 36.5 | 10 |
| OpenHardware Software got its agent moment. Hardware hasn't. OpenHardware is the missing piece: an agent that can see the board it programs. | 36.5 | 9 |
| Traces An AI agent can read your code but not your page, so it guesses why the button is dead. Traces gives it 17 WebMCP tools to search inside a r | 36.5 | 9.5 |
| Spatialize The floor plan that talks back. | 36.5 | 9.5 |
| Peira An instrument for the US benefits system. Earn one more dollar at the wrong income and a family can lose thousands in benefits at once. Peir | 36.5 | 9 |
| E-ternal × WebMCP An open-source ERP that works with your browser agent — same screen, same state. The agent drafts; only a human can approve. Built on WebMCP | 36 | 8 |
| Edgewright A redirect console a human and an agent drive together, over WebMCP. Test, audit and fix routing across 11 hosts, then prove a migration nev | 36 | 8 |
| ExceptionOSMCP The agent-operable command center for resolving business exceptions safely. | 36 | 8 |
| Recall Response Workbench — Recall the Right Items Find the right 37, protect the other 8,383, and put the irreversible decision behind a visible review control. | 36 | 8 |
| Bridge Studio A visible browser workspace that turns web workflows into typed, human-approved tools. | 36 | 9 |
| Crom’s Research Desk: WebMCP-Powered Agentic Publishing An end-to-end WebMCP workspace where humans and agents search, compare, draft, and publish together through one visible webpage—no separate | 36 | 9 |
| DungeonQ Enter the unknown. Exit with proof. | 36 | 9 |
| EvidenceBound WebMCP Authority Compiler A human approves one release, the AI gains the WebMCP release tool, and correction or revocation makes it disappear. | 36 | 9 |
| FableCut A browser video editor your agent can drive. FableCut exposes its whole timeline as WebMCP tools, you and your agent cut the same edit, live | 36 | 9 |
| FORGE — Human-Agent World Laboratory A human-governed WebMCP control plane where agents inspect, repair, test, and safely reverse changes to a shared game world. | 36 | 9 |
| Semantic City — Shared-World WebMCP Simulator A playable city-management sandbox where humans and agents share one live city. WebMCP exposes traffic, budgets, and utilities so every agen | 36 | 9 |
| AgentReady.js: Instant WebMCP for Every Website Turn any website into safe, structured WebMCP tools with a single script tag. Progressive enhancement for AI agents with built-in human-in-t | 36 | 9 |
| CanvasOps CanvasOps is a shared cloud architecture canvas where developers and AI design, analyze, and safely modify infrastructure together in real t | 36 | 9 |
Scores are the consolidated four-criterion sums from the analysis pipeline (each criterion 1–10, aggregate 4–40). 1,353 projects were re-scored (2026-09-05) after a video-evidence defect was corrected; those rows carry a single re-reviewer's confidence, the rest the mean of two blind Stage 1 reviewers. “—” means the reviewers' scores were withheld (they differ by more than 2) or the datum is unavailable; it never means zero. Open any row to read the reviewers' full reasoning.
No matches for that search or filter.