feat(discovery): add Agentic Resource Discovery (ARD) catalog for the lci MCP

Ship a ready-to-host ARD catalog (https://agenticresourcediscovery.org) so AI
agents can discover the lci MCP server by crawling a well-known JSON file:

- .well-known/ai-catalog.json — ARD catalog listing the lci MCP server, its 14
  tools as capabilities, and representative natural-language queries.
- .well-known/lci-mcp.json — the entry's config: stdio MCP (`command: lci`,
  `args: [mcp]`) plus install one-liners.
- docs/ARD-PUBLISHING.md — how to host (domain root / GitHub Pages / DNS TXT),
  the required HTTPS + JSON content-type + CORS, and which identifiers to change
  if served from a domain other than standardbeagle.com.

Identifiers are anchored to standardbeagle.com (did:web / urn:air); the doc
notes how to repoint them. lci's MCP is stdio-only, so the catalog references a
static config rather than a hosted endpoint.

Co-Authored-By: Claude Opus 4.8 <[email protected]>

Author: andylbrummer

.well-known/ai-catalog.json

@@ -0,0 +1,39 @@
+{
+  "specVersion": "1.0",
+  "host": {
+    "displayName": "Standard Beagle",
+    "identifier": "did:web:standardbeagle.com"
+  },
+  "entries": [
+    {
+      "identifier": "urn:air:standardbeagle.com:server:lci",
+      "displayName": "LCI — Lightning Code Index",
+      "type": "application/mcp-server+json",
+      "url": "https://standardbeagle.com/.well-known/lci-mcp.json",
+      "description": "Sub-millisecond semantic code search and code-intelligence MCP server. Indexes a repository with tree-sitter across 13 languages and answers search, symbol lookup, call-hierarchy, references, file outline, and git-change queries with ~80% less context than grep. Runs locally over stdio (`lci mcp`).",
+      "capabilities": [
+        "info",
+        "search",
+        "find_files",
+        "get_context",
+        "context",
+        "list_symbols",
+        "inspect_symbol",
+        "browse_file",
+        "index_stats",
+        "debug_info",
+        "semantic_annotations",
+        "side_effects",
+        "code_insight",
+        "git_analysis"
+      ],
+      "representativeQueries": [
+        "find where the authentication middleware is defined",
+        "show me all callers of the ServeHTTP function",
+        "search the codebase for functions that parse JSON",
+        "what does index_directory call and is it pure",
+        "list every struct type in the parser package"
+      ]
+    }
+  ]
+}

.well-known/lci-mcp.json

@@ -0,0 +1,14 @@
+{
+  "name": "lci",
+  "displayName": "LCI — Lightning Code Index",
+  "description": "Sub-millisecond semantic code search and code-intelligence MCP server (stdio). Index a repo once, then search content, look up symbols, trace call hierarchies, find references, outline files, and analyze git change — with ~80% less context than grep.",
+  "transport": "stdio",
+  "command": "lci",
+  "args": ["mcp"],
+  "homepage": "https://github.com/standardbeagle/lci-cpp",
+  "install": {
+    "linux_macos": "curl -fsSL https://raw.githubusercontent.com/standardbeagle/lci-cpp/main/install.sh | sh",
+    "windows": "irm https://raw.githubusercontent.com/standardbeagle/lci-cpp/main/install.ps1 | iex",
+    "npm": "npm install -g @standardbeagle/lci"
+  }
+}

docs/ARD-PUBLISHING.md

@@ -0,0 +1,47 @@
+# Publishing LCI for Agentic Resource Discovery (ARD)
+
+[ARD](https://agenticresourcediscovery.org/how_to_publish/) lets AI agents and
+chatbots discover MCP servers, skills, and tools by crawling a well-known JSON
+catalog. This repo ships a ready-to-host catalog for the LCI MCP server.
+
+## Files
+
+| File | Hosted as | Purpose |
+|------|-----------|---------|
+| [`.well-known/ai-catalog.json`](../.well-known/ai-catalog.json) | `https://<domain>/.well-known/ai-catalog.json` | ARD catalog: lists the LCI MCP server, its 14 tools, and example queries |
+| [`.well-known/lci-mcp.json`](../.well-known/lci-mcp.json) | the catalog entry's `url` | stdio MCP config (`command: lci`, `args: [mcp]`) + install one-liners |
+
+The catalog points at LCI's MCP server, which runs locally over stdio (`lci mcp`)
+— there is no hosted HTTP endpoint to keep alive, only these two static files.
+
+## Hosting (one of)
+
+ARD requires **HTTPS**, `Content-Type: application/json`, and
+`Access-Control-Allow-Origin: *` on the served files.
+
+1. **Domain root** (`standardbeagle.com`) — copy `.well-known/` to the web root.
+   Set the JSON content-type and the CORS header on that path.
+
+2. **GitHub Pages user site** (`standardbeagle.github.io`) — publish these files
+   from a `standardbeagle.github.io` repo. GitHub Pages serves `.well-known/`
+   and already sends `Access-Control-Allow-Origin: *`, so no extra config.
+
+3. **DNS TXT pointer** — if the catalog cannot live at the domain root, add a
+   TXT record `_catalog._agents.standardbeagle.com` whose value is the full
+   catalog URL (e.g. a GitHub Pages project path or raw URL).
+
+## When the host domain changes
+
+The identifiers are domain-anchored. If you serve from a domain other than
+`standardbeagle.com`, update in `.well-known/ai-catalog.json`:
+
+- `host.identifier` → `did:web:<domain>`
+- each `entries[].identifier` → `urn:air:<domain>:server:lci`
+- `entries[].url` → the hosted location of `lci-mcp.json`
+
+## Verify
+
+```sh
+curl -fsSL https://<domain>/.well-known/ai-catalog.json | python3 -m json.tool
+curl -fsSI https://<domain>/.well-known/ai-catalog.json | grep -i 'content-type\|access-control'
+```