MCP Server

Architecture audit,
inside your AI tool

Archiet's architecture-audit factory is available as a public MCP server. Connect it to Claude Desktop, Cursor, or any MCP-compatible tool and get consulting-grade architecture audits without leaving your IDE.

Streamable HTTP · JSON-RPC 2.0 · Stateless · No auth required

Connect in 60 seconds

One endpoint: https://archiet.com/api/mcp. Drop it into your MCP client config and your AI assistant gains architecture auditing.

Claude Desktop

// ~/Library/Application Support/Claude/claude_desktop_config.json
// (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)
{
  "mcpServers": {
    "archiet": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-fetch",
        "https://archiet.com/api/mcp"
      ]
    }
  }
}

Cursor

// .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)
{
  "mcpServers": {
    "archiet": {
      "url": "https://archiet.com/api/mcp"
    }
  }
}

Any HTTP-capable MCP client

// Streamable HTTP transport (stateless)
POST https://archiet.com/api/mcp
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/list",
  "params": {}
}

Available tools

Three tools exposed over tools/list. Call them from your AI assistant's tool-use interface.

submit_architecture_audit

Submit an architecture description and get back a consulting-grade audit: severity-ranked gaps, traceability across 22+ concerns, and a 30/60/90-day roadmap. Returns a shareable URL.

input { email, text?, documents? }output { token, share_url, audit_url }

get_audit_status

Fetch a previously submitted audit by token — status, inferred customer profile, report URLs.

input { token }output { request, share_url, audit_url }

get_audit_limits

Retrieve upload limits (max documents, bytes per file, total) before submitting large document sets.

input {}output { max_document_count, max_total_bytes, rate_limit }

Why an architecture audit as an MCP tool?

In-context analysis

Your AI assistant can submit the architecture you're discussing, get a structured audit, and reason over the findings — all in one conversation.

Deterministic, not hallucinated

The audit engine is rule-based, not LLM-driven. The same architecture in always produces the same findings out — reproducible and auditable.

Shareable reports

Every audit returns a shareable, unguessable URL your AI assistant can include in its response — forward it to your team or advisor.

Technical specification

Endpoint	https://archiet.com/api/mcp
Transport	Streamable HTTP (stateless POST)
Protocol	JSON-RPC 2.0 / MCP 2025-06-18
Auth	None — public endpoint
Rate limit	1 audit / 60 s / IP (for submit_architecture_audit)
Batch requests	Not supported (removed from MCP spec 2025-06-18)
Max body	~21 MB (upload cap + JSON overhead)
GET /api/mcp	405 — stateless server, no SSE stream
Analytics	None — no telemetry on request payloads

Frequently asked questions

What is the Archiet MCP server?

Archiet exposes its architecture-audit factory as a public Model Context Protocol (MCP) server at https://archiet.com/api/mcp. MCP is an open standard, developed by Anthropic, that lets AI tools — Claude Desktop, Cursor, GitHub Copilot extensions, and hundreds of others — call external services as structured tools. Once connected, your AI assistant can submit architecture descriptions to Archiet and receive a consulting-grade audit report without you leaving your IDE.

Is the MCP server free to use?

Yes. The audit lead magnet is anonymous by design — no account required. One audit per minute per IP applies. The result is a consulting-grade report at a shareable URL you can forward to your team or advisor.

What does "consulting-grade" mean?

Archiet's audit factory scores every submission against an 80-point gate: executive summary, severity-ranked findings, evidence references, a phased roadmap, and zero placeholder sections. Submissions that don't carry enough architectural detail to clear the gate are rejected with specific reasons rather than producing a low-quality report.

Does the MCP server use an LLM to analyse my architecture?

No. The analysis pipeline is deterministic — a rule-based engine scores each submitted architecture against 22+ concern dimensions without any LLM in the hot path. The determinism is why reports are reproducible and auditable. The LLM-boundary rule governing the Archiet codebase prohibits LLM calls in the analysis path.

What format should I use for the architecture description?

Plain English, Markdown, ArchiMate-export JSON, C4 diagrams pasted as text, ADRs, or any mix. The richer the description — components, data flows, compliance scope, integrations, scale assumptions — the higher-quality the audit. A minimal description (one paragraph about a web app) will be rejected if it scores below the consulting-grade floor.

Can I submit multiple documents?

Yes. The `documents` parameter accepts an array of `{ name, content }` objects — architecture.md, decisions.md, data-flow.md in one call. Total size caps are returned by the `get_audit_limits` tool.

Add architecture auditing to your AI tool

One endpoint. No account. Connect in 60 seconds.

Loading…

Architecture audit,
inside your AI tool

Streamable HTTP · JSON-RPC 2.0 · Stateless · No auth required

Connect in 60 seconds

One endpoint: https://archiet.com/api/mcp. Drop it into your MCP client config and your AI assistant gains architecture auditing.

Claude Desktop

// ~/Library/Application Support/Claude/claude_desktop_config.json
// (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)
{
  "mcpServers": {
    "archiet": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-fetch",
        "https://archiet.com/api/mcp"
      ]
    }
  }
}

Cursor

// .cursor/mcp.json (project) or ~/.cursor/mcp.json (global)
{
  "mcpServers": {
    "archiet": {
      "url": "https://archiet.com/api/mcp"
    }
  }
}

Any HTTP-capable MCP client

// Streamable HTTP transport (stateless)
POST https://archiet.com/api/mcp
Content-Type: application/json

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/list",
  "params": {}
}

Available tools

Three tools exposed over tools/list. Call them from your AI assistant's tool-use interface.

submit_architecture_audit

Submit an architecture description and get back a consulting-grade audit: severity-ranked gaps, traceability across 22+ concerns, and a 30/60/90-day roadmap. Returns a shareable URL.

input { email, text?, documents? }output { token, share_url, audit_url }

get_audit_status

Fetch a previously submitted audit by token — status, inferred customer profile, report URLs.

input { token }output { request, share_url, audit_url }

get_audit_limits

Retrieve upload limits (max documents, bytes per file, total) before submitting large document sets.

input {}output { max_document_count, max_total_bytes, rate_limit }

Why an architecture audit as an MCP tool?

In-context analysis

Your AI assistant can submit the architecture you're discussing, get a structured audit, and reason over the findings — all in one conversation.

Deterministic, not hallucinated

The audit engine is rule-based, not LLM-driven. The same architecture in always produces the same findings out — reproducible and auditable.

Shareable reports

Every audit returns a shareable, unguessable URL your AI assistant can include in its response — forward it to your team or advisor.

Technical specification

Endpoint	https://archiet.com/api/mcp
Transport	Streamable HTTP (stateless POST)
Protocol	JSON-RPC 2.0 / MCP 2025-06-18
Auth	None — public endpoint
Rate limit	1 audit / 60 s / IP (for submit_architecture_audit)
Batch requests	Not supported (removed from MCP spec 2025-06-18)
Max body	~21 MB (upload cap + JSON overhead)
GET /api/mcp	405 — stateless server, no SSE stream
Analytics	None — no telemetry on request payloads

Architecture audit,inside your AI tool

Connect in 60 seconds

Available tools

submit_architecture_audit

get_audit_status

get_audit_limits

Why an architecture audit as an MCP tool?

In-context analysis

Deterministic, not hallucinated

Shareable reports

Technical specification

Frequently asked questions

What is the Archiet MCP server?

Is the MCP server free to use?

What does "consulting-grade" mean?

Does the MCP server use an LLM to analyse my architecture?

What format should I use for the architecture description?

Can I submit multiple documents?

Add architecture auditing to your AI tool

Architecture audit,inside your AI tool

Connect in 60 seconds

Available tools

submit_architecture_audit

get_audit_status

get_audit_limits

Why an architecture audit as an MCP tool?

In-context analysis

Deterministic, not hallucinated

Shareable reports

Technical specification

Frequently asked questions

What is the Archiet MCP server?

Is the MCP server free to use?

What does "consulting-grade" mean?

Does the MCP server use an LLM to analyse my architecture?

What format should I use for the architecture description?

Can I submit multiple documents?

Add architecture auditing to your AI tool

Architecture audit,
inside your AI tool

Architecture audit,
inside your AI tool