fix: revert to knowledge layer positioning, add cold-start framing

CLAUDE.md

@@ -1,6 +1,6 @@
 # CodeCortex
 
-Codebase navigation and risk layer for AI agents. Pre-builds a map of architecture, dependencies, coupling, and risk areas so agents go straight to the right files.
+Persistent codebase knowledge layer for AI agents. Pre-builds architecture, dependency, coupling, and risk knowledge so agents skip the cold start and go straight to the right files.
 
 ## Stack
 - TypeScript, ESM (`"type": "module"`)

README.md

@@ -1,6 +1,6 @@
 # CodeCortex
 
-Codebase navigation and risk layer for AI agents. Pre-builds a map of architecture, dependencies, coupling, and risk areas so agents go straight to the right files.
+Persistent codebase knowledge layer for AI agents. Pre-builds architecture, dependencies, coupling, and risk knowledge so agents skip the cold start and go straight to the right files.
 
 [![CI](https://github.com/rushikeshmore/CodeCortex/actions/workflows/ci.yml/badge.svg)](https://github.com/rushikeshmore/CodeCortex/actions/workflows/ci.yml)
 [![npm version](https://img.shields.io/npm/v/codecortex-ai)](https://www.npmjs.com/package/codecortex-ai)
@@ -24,9 +24,9 @@ Every AI coding session starts with exploration — grepping, reading wrong file
 
 ## The Solution
 
-CodeCortex gives agents a pre-built map: architecture, dependencies, risk areas, hidden coupling. The agent goes straight to the right files and starts working.
+CodeCortex eliminates the cold start. It pre-builds codebase knowledge — architecture, dependencies, risk areas, hidden coupling — so agents skip the exploration phase and go straight to the right files.
 
-**CodeCortex finds WHERE to look. Your agent still reads the code.**
+**Not a middleware. Not a proxy. Just knowledge your agent loads on day one.**
 
 Tested on a real 6,400-file codebase (143K symbols, 96 modules):
 

package.json

@@ -1,7 +1,7 @@
 {
   "name": "codecortex-ai",
   "version": "0.5.0",
-  "description": "Codebase navigation and risk layer for AI agents — architecture, dependencies, coupling, and risk areas served via MCP",
+  "description": "Persistent codebase knowledge layer for AI agents — architecture, dependencies, coupling, and risk served via MCP",
   "type": "module",
   "bin": {
     "codecortex": "dist/cli/index.js"

src/mcp/server.ts

@@ -2,7 +2,7 @@
  * CodeCortex MCP Server
  *
  * Serves codebase knowledge to AI agents via Model Context Protocol.
- * 13 tools: 10 read (navigation + risk) + 3 write (knowledge creation).
+ * 13 tools: 8 read + 5 write (navigation, risk, memory).
  *
  * Usage:
  *   codecortex serve
@@ -28,7 +28,7 @@ export function createServer(projectRoot: string): McpServer {
   const server = new McpServer({
     name: 'codecortex',
     version: '0.5.0',
-    description: 'Codebase navigation and risk layer for AI agents. Pre-built map of architecture, dependencies, coupling, and risk areas.',
+    description: 'Persistent codebase knowledge layer for AI agents. Architecture, dependencies, coupling, risk, and cross-session memory.',
   })
 
   registerReadTools(server, projectRoot)