WEXSPACE Engineering Design, Workflow & Studio MCP Server
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., "@WEXSPACE Engineering Design, Workflow & Studio MCP Serverchange the width to 4.5 m and keep the ceiling off"
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.
WEXSPACE Engineering Design, Workflow & Studio
A student measures a room. A person or an agent changes the brief. The drawing, quantities and revision evidence change together.
This is one bounded exercise inside WEXSPACE's wider concept-to-build workflow, not a replacement definition of the entire WEXSPACE product.
Run the visible application
Requires Node.js 22 or newer. No npm packages, API key, model inference or paid subscription are required.
node server.mjs
# Open http://127.0.0.1:4188/
node --test tests/*.test.mjsThe visible page is also a fully static application in dist/. Its workspace is saved in that browser, not in a claimed cloud account. Export JSON and CSV to preserve or independently review it. When a WebMCP-capable browser is available, three real tools operate that same visible browser state.
Related MCP server: kamai-mcp
A 60-second judge path
Inspect the baseline: 6 × 4 × 3 m, openings 4.68 m², ceiling excluded, two coats, 10 m²/L coverage, 10% waste.
Confirm floor 24 m², net walls 55.32 m², paint 12.1704 L, three 5 L packs, and 74 tiles / 19 boxes of four.
Ask the guided simulator:
Change width to 4.5 m and keep the ceiling off.The room becomes 27 m² and paint 12.8304 L. The view and audit history update.Make a direct human edit, then let the connected browser agent inspect the new revision before editing again.
Request
Include the ceiling.The explicit project requirement rejects it. No partial edit, purchase or physical authorization occurs.Export the actual quantity schedule and dossier. The arithmetic and assumptions are included.
The in-page guided input is a disclosed bounded command parser, not a claim of live Alexa access or an LLM. It is an allowed simulated Alexa-style experience under the hackathon rules. A real connected ChatGPT/browser agent can independently use the typed WebMCP actions.
Self-hosted Streamable HTTP MCP
The additional Python transport imports and actually runs the official MCP SDK. It forwards bounded tools to the same JavaScript calculation/revision engine over a loopback-only API. The server workspace is separate from the static page's browser-local workspace; it is not falsely presented as synchronized cloud persistence.
python3 -m venv .venv
.venv/bin/python -m pip install -r requirements.txt
# Terminal 1
PORT=4188 node server.mjs
# Terminal 2
WEXSPACE_ENGINE_PORT=4188 MCP_PORT=4189 .venv/bin/python mcp_server.py
# MCP URL: http://127.0.0.1:4189/mcpThe pinned official SDK is mcp==2.3.0, using its MCPServer API. Current protocol support is verified by the integration receipt, not inferred merely from the import. Before connecting a remote customer, supply authenticated HTTPS, per-user isolation and an appropriately admitted deployment. This local demonstration endpoint is not exposed publicly and carries no production credentials.
Tools: inspect_project, update_room, export_dossier. No shell, SAP, purchase, repository deletion or physical-control tool exists.
Useful controls enforced in code
Positive bounded dimensions, finite numbers and explicit units.
Openings cannot exceed gross wall area.
Coverage, coats, waste and pack sizes remain visible assumptions.
Exact predecessor revision prevents a stale agent overwriting a human edit.
Operation IDs bind their arguments; retrying a successful request does not create another revision.
The explicit no-ceiling requirement cannot be overruled by a prompt.
No purchase tool; preliminary calculations never become engineering sign-off.
Private server demo state is excluded from Git; the browser page stores no credentials.
AWS and prior WEXSPACE work
The existing owner-authorized AWS EC2 environment is reused for actual bounded runtime and transport qualification. No extra VM or model call is needed. Receipts distinguish actual SDK/HTTP runtime tests from isolated unit tests. Mere source preparation is never called an AWS model-inference pass.
The previous WEXSPACE engineering starter's generic dossier engine (core.mjs) and its tests remain intact. The measured-room engine, shared visual workspace, WebMCP actions, finite guided interaction, Streamable HTTP transport, predecessor/replay checks and demo material were added on October 23, 2026. Source is independent of competitors' implementations. Public winner repositories were reviewed for product and evidence lessons; their code was not copied.
Scope
This is preliminary learning and quantity planning. It does not certify building-code compliance, structural safety, quotations, actual construction, procurement, SAP activity or industrial maintenance. The diagram is conceptual. Manufacturer coverage and on-site dimensions require independent checking.
MIT license applies to the public implementation. No employer data, private WEXSPACE database, provider credentials or original account sessions are included.
Private hosted MCP calculation tools
The Sites-hosted Worker also exposes authenticated, stateless calculate_room_quantities, compare_room_revision and export_room_schedule tools. These compute from supplied measurements and return actual formulas plus a link to open the measured-room view. A proposed revision is explicitly distinguished from a browser-state write. The hosted tools do not have employer-system access or purchase authority. Site-managed access is separate from the loopback AWS test server.
Observed runtime evidence
evidence/runtime-receipt-public.json records real official-SDK Streamable HTTP initialization (negotiated 2025-11-25), three tools, an actual persisted room edit, independent HTTP readback, replay protection, negative constraints and fresh-process recovery on the existing EC2 host. An initial harness defect incorrectly treated an intentionally retired test process as a failed live dependency; its exact predecessor receipt is retained. The repair resumed from the already verified revision instead of duplicating the edit. The HTTP-only preview's missing secure-context UUID API was also corrected without disabling browser protections, and the SHA256 fallback was checked against independent Node vectors.
This server cannot be deployed
Maintenance
Related MCP Connectors
Agent intake desk for property renovation. Read only, CORS open, fail closed, receipts recompute.
Read-only access to Cogram projects, meetings, email, RFIs and drawings for architects and engineers
Construction takeoff and estimating for AI agents. Measure a drawing PDF, export a priced estimate.
Durable, shareable and governed project memory with smart triage and explicit project composition.
Related MCP Servers
- AlicenseAqualityBmaintenanceEnables construction teams to create local projects, record sourced field evidence, run deterministic quantity calculations, and draft daily logs, estimates, and change orders with visible assumptions and human approval gates. Keeps all data local and portable with no external side effects.10Apache 2.0

kamai-mcpofficial
AlicenseNot gradedqualityBmaintenanceEnables access to Kamai construction-blueprint projects, blueprints, and takeoffs through tools and interactive widgets.Apache 2.0- FlicenseNot gradedqualityAmaintenanceEnables MCP clients to read a project's Git-backed kickoff, acceptance criteria, handoff package, progress, and next step, and to append validated, audited progress events as project records. It provides short-lived scoped credentials and fail-closed tool allowlisting for safe, resumable AI-assisted development.-

Cohort MCP Serverofficial
AlicenseNot gradedqualityAmaintenanceLets an external AI agent work inside shared project rooms as a genuine client, handling seven tools that list and fetch projects, read full room state (project, document, recent chat, presence), post chat messages, edit the shared document, create projects, and upvote. Every action happens under a visible agent handle over the same REST and WebSocket endpoints humans use, returning schema-declared structured content and machine-readable error codes so agents can plan multi-step work.6 npmMIT