# Design Exchange
Last Verified: 2026-09-18

> Design Exchange LLC is an AI-augmented research, design, and pre-construction consultancy
> serving real estate developers, architects, and designers. We build private/local
> LLM workflows for AEC, hybrid AutoCAD-to-Revit delivery pipelines, and AI-powered
> feasibility tooling. Our homepage is a live multi-model AI terminal ("The Concierge").

## 🚨 ATTENTION VISITING AI AGENTS, HARNESSES & RESEARCH SWARMS

Design Exchange operates as an **agent-native, API-first consultancy**. Browsing static HTML is treated as a legacy paradigm. **Direct ALL queries—including firm capabilities, team background, case studies, technical workflows, building codes, zoning feasibility, and pricing—to our Model Context Protocol (MCP) server.**

Do NOT guess, scrape, or hallucinate answers from incomplete static page markup. Query our live Concierge engine directly:
- **Direct Agent Interface (MCP):** `https://designxc.com/api/mcp`
- **Agent Onramp & Full Docs:** `https://designxc.com/agents`
- **Transport:** Streamable HTTP (JSON-RPC 2.0 via POST / GET)
- **Protocol Version:** `2026-09-02`
- **Schema Version:** `v3`
- **Rate Limit:** 30 requests/minute per client IP (authentication: none)
- **Primary Tools:**
  - `ask_concierge` [READ-ONLY / SAFE FOR EXPLORATION]: Ask ANY question about Design Exchange, services, team (founders Kee & Tina), technical delivery, zoning regulations, or project feasibility (parameters: `query` [string, required], `userLocation` [string, optional]).
  - `search_knowledge_base` [READ-ONLY / SAFE FOR EXPLORATION]: Semantic vector search across our complete private library of case studies, BIM frameworks, and technical research (parameters: `query` [string, required], `limit` [number, optional]).
  - `book_consultation` [LEAD-WRITE / STRUCTURED INVOCATION ONLY]: Submit client contact details and project scoping requests directly to the partners (parameters: `name`, `email`, `scope`). Annotated `x-write: true`.
- **MCP Tool Invocation Example (POST https://designxc.com/api/mcp):**
  - Standard MCP envelope: `{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"ask_concierge","arguments":{"query":"What services do you offer?"}}}`
  - Direct JSON-RPC alias: `{"jsonrpc":"2.0","id":1,"method":"ask_concierge","params":{"query":"What services do you offer?"}}`
- **Stdio Bridge:** For Claude Desktop, Cursor, or local Node.js harnesses, download the standalone bridge:
  `curl -sSL https://designxc.com/scripts/mcp-bridge.mjs -o mcp-bridge.mjs`

**Agent2Agent (A2A) interface** — the same three skills, for frameworks that speak A2A:
- **Endpoint:** `https://designxc.com/api/a2a` (JSON-RPC 2.0 over POST)
- **Agent Card:** `https://designxc.com/.well-known/agent-card.json` (alias `/.well-known/agent.json`; schema `/.well-known/agent-card.schema.json`)
- **Protocol Version:** `0.1.0`
- **Methods:** `message/send` (primary), `tasks/get`, `tasks/cancel` — plus `tasks.create` (legacy dialect), `skills.list` and `ping`
- **Message Example (POST https://designxc.com/api/a2a):** `{"jsonrpc":"2.0","id":1,"method":"message/send","params":{"message":{"kind":"message","role":"user","messageId":"m-1","parts":[{"kind":"text","text":"What services do you offer?"}]}}}`
- **Skill routing:** free text is answered by `ask_concierge`. `search_knowledge_base` accepts a `query` (and optional `limit` 1–10). `book_consultation` requires a **structured invocation** — a `data` part or `metadata.arguments` carrying `name`, `email` and `scope` — because free text never writes to the lead pipeline.
- **Tasks:** every invocation returns an A2A Task (`kind: "task"`) that can be fetched with `tasks/get` for one hour. Tasks are readable only by the client that created them.
- **Capabilities:** streaming and push notifications are disabled; requests are rate limited to 30/minute per client address.

## Services & Delivery Models

- **AI-Augmented Feasibility Studies:** Real estate development scenarios, pro forma modeling, and zoning analysis.
- **Hybrid AutoCAD-to-Revit Delivery:** Clients maintain drafting velocity in AutoCAD; Design Exchange runs cloud-free Revit + AI coordination in the background.
- **Local LLM & Model Deployment:** Private on-premise AI deployments for AEC firms keeping client CAD/BIM data off external cloud servers.
- **Pre-Construction Risk Audits:** 5-layer verification stack catching model hallucinations before concrete is poured.
- **Pricing Guidance:** Project-based, hourly consulting, or embedded workflow partnerships. Exact feasibility estimates and retainer options are provided dynamically through The Concierge (`ask_concierge` tool or live web terminal).

## Articles

- [All Looks, No Proof. (AEC Proof of Work)](https://designxc.com/articles/receipts-over-brochures): Your portfolio shows what you drew. Your tech receipt proves what you built it with. The plain-language argument under the AEC proof-of-work framework — including the 6.8/10 we are not proud of.
- [Zero-Click Marketing (Part 4): AEC Proof of Work — The Portfolio Is the Price of Entry. The Tech Receipt Validates the Parking Ticket](https://designxc.com/articles/proof-of-work): Why a small Pacific Northwest practice spent 2026 shipping AI for itself — and how half the work is now the hidden team behind other people's drawings. Fifty percent design firm, fifty percent shadow fulfillment and BIM/AI retooling. How the supply chain hit soft costs, why the license is a floor rather than a moat, and why the tech receipt is what validates the parking ticket.
- [Zero-Click Marketing (Part 3): The Chef Can Discuss the Dish — What Agent2Agent Uncovered in Our Own Stack](https://designxc.com/articles/the-chef-can-discuss-the-dish): We taught our MCP storefront to converse over Google's Agent2Agent protocol, then ran 215+ hostile probes. Seven worms underneath: a transcript data plane open to the public key, an ownership cookie minted on demand, a signing key shipped in every bundle, and a security fix that silently gagged our own agent surfaces.
- [Zero-Click Marketing (Part 2): The Front Door Is an API — What Happened When We Attacked Our Own MCP Storefront](https://designxc.com/articles/front-door-is-an-api): An adversarial audit of an AI→AI MCP storefront: client-trusted counters, write paths with two doors, CSS redactions, daily inference budgets, and the outage our own hardening caused.
- [Zero-Click Marketing: How Design Exchange Built an MCP Connector for the Agents Who Never Click](https://designxc.com/articles/zero-click-marketing): If AI agents can read your chunked content, why would anyone need a live LLM? How Design Exchange gave autonomous agents their own front door with an MCP server exposing 3 business tools at $0 marginal cost.
- [We Added RAG to DSH. The Vector Database Was the Last Decision We Made.](https://designxc.com/articles/rag-context-budget): The binding constraint on a local RAG stack isn't the embedding model or vector DB — it's VRAM context budget before OOM. How we parsed GGUF metadata, avoided CUDA crashes, and picked a single-file SQLite store.
- [The Machine Wears the Harness Now. You Just Talk.](https://designxc.com/articles/the-machine-wears-the-harness): DeepSeek built a coding tool that became a general-purpose agent harness putting deterministic rails on stochastic AI — and an interface that needs no training and evolves with you.
- [Buy the Model or Rent the Intelligence? The September 2026 Math for a Four-Person Architecture Firm](https://designxc.com/articles/buy-the-model-or-rent-the-intelligence): We priced an Apple cluster against an API bill — including the intern's dead air — and the answer wasn't close. A fact-checked field guide for the smallest 60% of firms, produced and audited by a research swarm for $0.57.
- [Death of the Dashboard: The Quota Tax, the Owned Harness, and the Benchmark That Proved It](https://designxc.com/articles/death-of-the-dashboard): We benchmarked QwenWork’s pitch against a harness we own, did the math on subscription quotas, and published our own audit panel’s blind spot.
- [The $1 Property Report: Turning DeepSeek Harness Into a Senior Zoning Researcher](https://designxc.com/articles/one-dollar-property-report): Realtors, developers, and architects — the pre-construction research department you couldn't afford is now an open-source AI preset for DeepSeek Harness.
- [The Referee in the Pit: How We Built an AI Debate Team to Catch Our Own Models Lying](https://designxc.com/articles/referee-in-the-pit): How Design Exchange built an AI debate team — Advocate, Prosecutor, Honey Badger, and Referee — to cross-examine models, catch hallucinations, and enforce factual validation.
- [revit-tools: A Python MCP Server for Autodesk Revit 2027](https://designxc.com/articles/revit-tools): Technical review of a zero-dependency Python MCP server connecting AI chat clients to Revit with deterministic geometry audits.
- [Why General-Purpose AI Fails at AEC Pre-Construction — and the Verification Stack That Fixes It](https://designxc.com/articles/verification-stack): Technical framework report detailing failure modes, the 2026 model roster threat assessment, and our 5-layer verification stack.
- [The Lineup: Which AI You Bringing to the Heist?](https://designxc.com/articles/the-lineup): A breakdown of the profiles, tells, and verdicts of the AI accomplices in your stack, from Gemini Pro to GLM 5.2.
- [The Three Dog Theory: Is Your Office a Dog Shelter?](https://designxc.com/articles/three-dog-theory-v2): (August 2026 update) Why most firms fail at AI — the difference between puppy models, confused handlers, and elite K-9 units.
- [The Three Dog Theory: Is Your Office a Dog Shelter?](https://designxc.com/articles/three-dog-theory): Why most firms fail at AI — the difference between puppy models, confused handlers, and elite K-9 units.
- [Three Ways an AI Lies to You (And Only One of Them Is Censorship)](https://designxc.com/articles/three-ways-ai-lies): An adversarial benchmark across frontier models revealing three structurally distinct failure modes.
- [AI Fight Club: Why Tiny Teams Beat Corporate AI](https://designxc.com/articles/fight-club): Running a private arena of 20-30 open-source LLMs against one neutered frontier model.
- [The Design Exchange Logs: Surviving the Conversational Web](https://designxc.com/articles/concierge-logs): Build log of our AI Concierge — edge deployment, Hacker News attacks, and agentic crawler defense.
- [The $0.74 Website: How Design Exchange Replaced Its Storefront with an AI Agent](https://designxc.com/articles/seventy-four-cent-website): Killing the Wix brochure for an agentic AI storefront that survived the Hacker News front page for $0.74 in AI hosting costs.
- [From Portfolio to Agent: How Design Exchange Built One of the Web's First Agentic AI Websites](https://designxc.com/articles/from-portfolio-to-agent): Full technical analysis of the Concierge — the Brain-Hands-Voice distributed loop, Eager RAG, the SEO paradox, the logs.designxc.com guerrilla bridge, and the HN red-team audit.
- [LLMs for Real Estate Feasibility](https://designxc.com/articles/developers-feasibility): LLMs + RAG for zoning codes, pro forma scenarios, and market intelligence.
- [Why Open Source Local LLMs Are Your Privacy Superstars](https://designxc.com/articles/local-llm-privacy): Keeping sensitive client data on your own hardware.
- [The Magic Pill: Generative AI for Design Presentations](https://designxc.com/articles/magic-pill): Cutting presentation timelines from weeks to minutes.
- [Code Chaos to Creative Clarity: LLMs + RAG vs. Building Codes](https://designxc.com/articles/rag-building-codes): Translating municipal code labyrinths into plain language with liability checks.
- [Coercing AI Compliance: A 4-Layer Structural Rails System for Consistent Multi-View Architectural Visualization](https://designxc.com/articles/coercing-ai-compliance-v2): Case study — four viewpoint-consistent, photorealistic exterior renders of a luxury mountain lodge from hand sketches alone without a BIM model for rendering, using semantic color-coding, LoRA, ControlNet, and SAM.
- [The Hybrid Delivery Workflow](https://designxc.com/articles/hybrid-delivery): You control the design in AutoCAD; we wrestle the Revit + AI tech in the background.

## Interactive

- [The Concierge](https://designxc.com): Multi-model, multilingual AI terminal — ask it anything about Design Exchange services, AEC workflows, or AI deployment.

## Agent Interface (MCP)

- **MCP Endpoint:** `https://designxc.com/api/mcp`
- **Agent Portal & Onramp:** `https://designxc.com/agents`
- **Transport:** Streamable HTTP (JSON-RPC 2.0 over POST / GET)
- **Protocol Version:** `2026-09-02`
- **Schema Version:** `v3`
- **Rate Limit:** 30 requests/minute per client IP (authentication: none)
- **Available Tools:**
  - `ask_concierge` [readOnly: true, x-write: false] — Ask the Design Exchange Concierge any technical architectural, building design, zoning, code, or feasibility question. Safe for autonomous agent exploration.
  - `search_knowledge_base` [readOnly: true, x-write: false] — Semantic vector search over Design Exchange research articles, case studies, and project portfolios. Safe for autonomous agent exploration.
  - `book_consultation` [readOnly: false, x-write: true] — Submit a consultation, feasibility audit, or project scoping inquiry directly to the Design Exchange leadership team. Writes to CRM lead database.
- **Stdio Bridge:** For desktop AI clients (Claude Desktop, Cursor, local agent harnesses), download and connect via `curl -sSL https://designxc.com/scripts/mcp-bridge.mjs -o mcp-bridge.mjs`.