neuronlayer 0.2.7 → 0.2.10

package/README.md

@@ -48,9 +48,9 @@ This registers your project and configures Claude Desktop, Claude Code, OpenCode
 | Tool | Setup |
 |------|-------|
 | Claude Desktop | `neuronlayer init` (auto) |
-| Claude Code (CLI) | [`neuronlayer init` (auto)](./documentation/CLAUDE-CODE-SETUP.md) |
+| Claude Code (CLI) | [`neuronlayer init` (auto)](./doc/CLAUDE-CODE-SETUP.md) |
 | Cursor | `neuronlayer init` (auto) |
-| OpenCode | [`neuronlayer init` (auto)](./documentation/OPENCODE-SETUP.md) |
+| OpenCode | [`neuronlayer init` (auto)](./doc/OPENCODE-SETUP.md) |
 | Any MCP Client | Manual config |
 | **Any tool (HTTP)** | `neuronlayer serve` |
 

package/doc/CLAUDE-CODE-SETUP.md

@@ -0,0 +1,504 @@
+# NeuronLayer + Claude Code Setup Guide
+
+This guide explains how to configure NeuronLayer as an MCP server for [Claude Code](https://code.claude.com/) (Anthropic's CLI tool).
+
+---
+
+## Prerequisites
+
+1. **Node.js 18+** installed
+2. **Claude Code** installed (`npm install -g @anthropic-ai/claude-code`)
+3. **NeuronLayer** built (`npm run build`)
+
+---
+
+## Quick Setup (Recommended)
+
+The easiest and most reliable way to configure NeuronLayer for Claude Code is using our automated initialization command.
+
+### Step 1: Initialize NeuronLayer
+
+Run this command in the root of your project:
+
+```bash
+npx neuronlayer init .
+```
+
+This will automatically configure `.mcp.json` in your project with the exact absolute paths and platform-specific commands required for Claude Code to connect smoothly to NeuronLayer.
+
+### Step 2: Verify Connection
+
+Start Claude Code and check your servers:
+
+```bash
+# Inside Claude Code
+/mcp
+```
+
+You should see:
+```
+MCP Servers
+  neuronlayer (stdio)
+    Status: connected
+```
+
+---
+
+## Configuration Methods (Advanced)
+
+If you prefer not to use the automated `npx neuronlayer init` command, Claude Code provides several ways to configure MCP servers manually.
+
+### Method 1: CLI Commands
+
+Claude Code provides CLI commands to manage MCP servers:
+
+```bash
+# Add a server
+claude mcp add --transport stdio neuronlayer -- node /absolute/path/to/dist/index.js --project .
+
+# List all servers
+claude mcp list
+
+# Get server details
+claude mcp get neuronlayer
+
+# Remove a server
+claude mcp remove neuronlayer
+```
+
+**Note:** On Windows natively, you MUST wrap the node execution with `cmd /c` to prevent stdin/stdout hanging issues:
+```bash
+claude mcp add --transport stdio neuronlayer -- cmd /c node C:\absolute\path\to\dist\index.js --project .
+```
+
+### Method 2: Project Configuration (`.mcp.json`)
+
+If you want to create `.mcp.json` in your project root manually:
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "node",
+      "args": ["/absolute/path/to/neuronlayer/dist/index.js", "--project", "."],
+      "env": {}
+    }
+  }
+}
+```
+
+*For Windows, use `["cmd", "/c", "node", "C:\\absolute\\path\\to\\dist\\index.js", "--project", "."]` in the command and args fields as appropriate depending on your node integration.*
+
+---
+
+## Configuration Scopes
+
+| Scope | Storage Location | Use Case |
+|-------|------------------|----------|
+| `local` (default) | `~/.claude.json` | Personal, current project only |
+| `project` | `.mcp.json` | Shared with team via version control |
+| `user` | `~/.claude.json` | Personal, available across all projects |
+
+```bash
+# The automated init command uses Project scope by default
+npx neuronlayer init .
+```
+
+---
+
+## Platform-Specific Setup
+
+### Windows (Native)
+
+On native Windows (not WSL), use the `cmd /c` wrapper for proper execution:
+
+```bash
+# Using CLI
+claude mcp add --transport stdio neuronlayer -- cmd /c node C:\path\to\neuronlayer\dist\index.js --project .
+```
+
+**In `.mcp.json`:**
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "cmd",
+      "args": ["/c", "node", "C:\\path\\to\\neuronlayer\\dist\\index.js", "--project", "."],
+      "env": {}
+    }
+  }
+}
+```
+
+### Windows (WSL)
+
+WSL works like Linux:
+
+```bash
+claude mcp add --transport stdio neuronlayer -- node /path/to/neuronlayer/dist/index.js --project .
+```
+
+### macOS / Linux
+
+```bash
+claude mcp add --transport stdio neuronlayer -- node /path/to/neuronlayer/dist/index.js --project .
+```
+
+**In `.mcp.json`:**
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "node",
+      "args": ["/Users/you/neuronlayer/dist/index.js", "--project", "."],
+      "env": {}
+    }
+  }
+}
+```
+
+---
+
+## After npm Publish
+
+Once NeuronLayer is published to npm, setup becomes simpler:
+
+### CLI Command
+
+```bash
+# Install globally
+npm install -g neuronlayer
+
+# Add to Claude Code
+claude mcp add --transport stdio neuronlayer -- neuronlayer --project .
+```
+
+### Using npx
+
+```bash
+# Without installation
+claude mcp add --transport stdio neuronlayer -- npx -y neuronlayer --project .
+```
+
+**Windows (with npx):**
+
+```bash
+claude mcp add --transport stdio neuronlayer -- cmd /c npx -y neuronlayer --project .
+```
+
+### Project `.mcp.json`
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "npx",
+      "args": ["-y", "neuronlayer", "--project", "."],
+      "env": {}
+    }
+  }
+}
+```
+
+---
+
+## Environment Variables
+
+### NeuronLayer Configuration
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `NEURONLAYER_MAX_TOKENS` | 6000 | Maximum tokens for context assembly |
+| `NEURONLAYER_DEBUG` | false | Enable debug logging |
+
+```bash
+claude mcp add --transport stdio \
+  --env NEURONLAYER_MAX_TOKENS=8000 \
+  --env NEURONLAYER_DEBUG=true \
+  neuronlayer -- node dist/index.js --project .
+```
+
+### Claude Code MCP Settings
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `MCP_TIMEOUT` | 60000 | Server startup timeout (ms) |
+| `MAX_MCP_OUTPUT_TOKENS` | 25000 | Max tokens from MCP tools |
+
+```bash
+# Start Claude Code with custom MCP settings
+MCP_TIMEOUT=120000 MAX_MCP_OUTPUT_TOKENS=50000 claude
+```
+
+---
+
+## Available MCP Tools
+
+Once connected, Claude Code can use these NeuronLayer tools:
+
+### Core Tools
+
+| Tool | Description |
+|------|-------------|
+| `get_context` | Get relevant codebase context for a query |
+| `search_codebase` | Semantic search across all files |
+| `record_decision` | Save an architectural decision |
+| `get_file_context` | Get content of a specific file |
+| `get_project_summary` | Overview of project structure |
+
+### Symbol & Dependency Tools
+
+| Tool | Description |
+|------|-------------|
+| `get_symbol` | Find functions, classes, types by name |
+| `get_dependencies` | See imports/exports for a file |
+| `get_file_summary` | Get compressed file summary (10x smaller) |
+
+### Living Documentation Tools
+
+| Tool | Description |
+|------|-------------|
+| `generate_docs` | Generate documentation for a file or architecture |
+| `get_architecture` | Get project architecture overview |
+| `get_component_doc` | Get detailed documentation for a component |
+| `get_changelog` | Get changelog of recent changes |
+| `validate_docs` | Check for outdated docs |
+| `what_happened` | Query recent project activity |
+
+### Context Health Tools
+
+| Tool | Description |
+|------|-------------|
+| `get_context_health` | Check context health and drift score |
+| `trigger_compaction` | Manually trigger context compaction |
+| `mark_critical` | Mark content as critical (never compress) |
+
+### Confidence & Intelligence Tools
+
+| Tool | Description |
+|------|-------------|
+| `get_confidence` | Get confidence score for code suggestion |
+| `check_conflicts` | Check if code conflicts with past decisions |
+| `what_changed` | Query what changed in the codebase |
+| `why_broke` | Diagnose why something broke |
+| `suggest_fix` | Get fix suggestions for an error |
+
+### Architecture Tools
+
+| Tool | Description |
+|------|-------------|
+| `validate_pattern` | Validate code against established patterns |
+| `suggest_existing` | Find existing functions that match intent |
+| `learn_pattern` | Teach a new pattern to the system |
+| `list_patterns` | List all learned patterns |
+
+### Test Awareness Tools
+
+| Tool | Description |
+|------|-------------|
+| `get_related_tests` | Get tests related to a file or function |
+| `check_tests` | Check if a code change would break tests |
+| `suggest_test_update` | Get suggested test updates for a change |
+| `get_test_coverage` | Get test coverage for a file |
+
+---
+
+## Example Usage
+
+Once configured, try these prompts in Claude Code:
+
+```
+> Search for authentication code in this project
+
+> What architectural decisions have been made?
+
+> Show me the project summary
+
+> Find the function that handles user login
+
+> What files import the database module?
+
+> What changed in the last 24 hours?
+
+> Why might the login test be failing?
+
+> Record a decision: We chose SQLite for local storage because it's embedded and requires no separate server
+
+> Get the architecture overview of this project
+
+> Check if this code follows our patterns: async function fetchUser(id) { ... }
+```
+
+---
+
+## Troubleshooting
+
+### "Connection closed" on Windows
+
+Ensure you're using the `cmd /c` wrapper:
+
+```bash
+# Wrong
+claude mcp add --transport stdio neuronlayer -- node dist/index.js --project .
+
+# Correct
+claude mcp add --transport stdio neuronlayer -- cmd /c node dist/index.js --project .
+```
+
+### "Server not found" or "ENOENT"
+
+1. Verify the path to `dist/index.js` is correct and absolute
+2. Ensure NeuronLayer is built: `npm run build`
+3. Check Node.js is in your PATH
+
+### "MCP server not connected"
+
+1. Run `claude mcp list` to check configuration
+2. Run `/mcp` inside Claude Code to see status
+3. Check for errors in the server output
+
+### "Configuration is invalid"
+
+Ensure your JSON is valid and `command` is a string (not array):
+
+```json
+// Correct
+{
+  "type": "stdio",
+  "command": "node",
+  "args": ["dist/index.js", "--project", "."]
+}
+
+// Wrong
+{
+  "type": "stdio",
+  "command": ["node", "dist/index.js"]
+}
+```
+
+### Server Takes Too Long to Start
+
+Increase the MCP timeout:
+
+```bash
+MCP_TIMEOUT=120000 claude
+```
+
+Or set in `.mcp.json`:
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "node",
+      "args": ["dist/index.js", "--project", "."],
+      "env": {},
+      "timeout": 120000
+    }
+  }
+}
+```
+
+### View Debug Logs
+
+```bash
+# Run Claude Code with debug logging
+claude --log-level DEBUG
+```
+
+---
+
+## Importing from Claude Desktop
+
+If you already have NeuronLayer configured in Claude Desktop:
+
+```bash
+# Import all servers from Claude Desktop
+claude mcp add-from-claude-desktop
+
+# Select neuronlayer from the list
+
+# Verify import
+claude mcp list
+```
+
+---
+
+## Data Storage
+
+NeuronLayer stores data locally:
+
+```
+~/.neuronlayer/
+├── registry.json          # Project registry
+└── projects/
+    └── {project-name}-{hash}/
+        ├── neuronlayer.db  # SQLite database
+        └── tier1.json      # Working context
+```
+
+---
+
+## Comparison: Claude Desktop vs Claude Code
+
+| Feature | Claude Desktop | Claude Code |
+|---------|----------------|-------------|
+| Config Location | `~/.claude/claude_desktop_config.json` | `~/.claude.json` or `.mcp.json` |
+| Add Server | Manual JSON edit | CLI commands or JSON |
+| Project Scope | No | Yes (`.mcp.json`) |
+| User Scope | Yes | Yes |
+| Local Scope | N/A | Yes (default) |
+| Env Var Expansion | No | Yes (`${VAR}` syntax) |
+
+---
+
+## Links
+
+- [Claude Code Documentation](https://code.claude.com/docs/)
+- [Claude Code MCP Guide](https://code.claude.com/docs/en/mcp)
+- [Claude Code Settings](https://code.claude.com/docs/en/settings)
+- [MCP Protocol Specification](https://modelcontextprotocol.io/)
+- [NeuronLayer Documentation](./INDEX.md)
+
+---
+
+## Quick Reference
+
+### Minimum Setup
+
+```bash
+# One-liner setup
+claude mcp add --transport stdio neuronlayer -- node /path/to/neuronlayer/dist/index.js --project .
+```
+
+### Full `.mcp.json` Example
+
+```json
+{
+  "mcpServers": {
+    "neuronlayer": {
+      "type": "stdio",
+      "command": "node",
+      "args": [
+        "/path/to/neuronlayer/dist/index.js",
+        "--project",
+        "."
+      ],
+      "env": {
+        "NEURONLAYER_MAX_TOKENS": "8000"
+      }
+    }
+  }
+}
+```
+
+---
+
+*Last updated: February 2026*

package/doc/OPENCODE-SETUP.md

@@ -0,0 +1,425 @@
+# NeuronLayer + OpenCode Setup Guide
+
+This guide explains how to configure NeuronLayer as an MCP server for [OpenCode](https://opencode.ai).
+
+---
+
+## Prerequisites
+
+1. **Node.js 18+** installed
+2. **OpenCode** installed (`npm install -g opencode`)
+3. **NeuronLayer** built (`npm run build`)
+
+---
+
+## Quick Setup (Recommended)
+
+The easiest and most reliable way to configure NeuronLayer for OpenCode is using our automated initialization command.
+
+### Step 1: Initialize NeuronLayer
+
+Run this command in the root of your project:
+
+```bash
+npx neuronlayer init .
+```
+
+This will automatically detect your operating system and generate the correct `opencode.json` configuration file with the exact absolute paths required for OpenCode to connect to NeuronLayer.
+
+### Step 2: Start OpenCode
+
+```bash
+opencode
+```
+
+### Step 3: Verify Connection
+
+Inside OpenCode, type:
+```
+/mcp
+```
+
+You should see:
+```
+MCP Servers
+  neuronlayer (local)
+    Status: connected
+```
+
+Or from your terminal:
+```bash
+opencode mcp list
+```
+
+---
+
+## Manual Configuration (Advanced)
+
+If you prefer to configure OpenCode manually, create an `opencode.json` file in your project root.
+
+**IMPORTANT:** You must use **absolute paths** for reliable operation, and on Windows you must use `cmd` to wrap the Node execution.
+
+**Windows Example:**
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["cmd", "/c", "node", "C:\\Users\\YOUR_USERNAME\\path\\to\\neuronlayer\\dist\\index.js", "--project", "C:\\Users\\YOUR_USERNAME\\path\\to\\your\\project"],
+      "enabled": true
+    }
+  }
+}
+```
+
+**macOS/Linux Example:**
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["node", "/home/user/neuronlayer/dist/index.js", "--project", "/home/user/myproject"],
+      "enabled": true
+    }
+  }
+}
+```
+
+---
+
+## Global Setup
+
+For system-wide availability, create the config in your home directory:
+
+**Windows:** `%USERPROFILE%\.config\opencode\opencode.json`
+**macOS/Linux:** `~/.config/opencode/opencode.json`
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["node", "/absolute/path/to/neuronlayer/dist/index.js", "--project", "."],
+      "enabled": true
+    }
+  }
+}
+```
+
+---
+
+## After npm Publish
+
+Once NeuronLayer is published to npm, setup becomes simpler:
+
+### Using npx
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["npx", "-y", "neuronlayer", "--project", "."],
+      "enabled": true
+    }
+  }
+}
+```
+
+### Global Install
+
+```bash
+npm install -g neuronlayer
+```
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["neuronlayer", "--project", "."],
+      "enabled": true
+    }
+  }
+}
+```
+
+---
+
+## Configuration Reference
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `type` | `"local"` \| `"remote"` | Yes | Server type |
+| `command` | `string[]` | Yes (local) | Command array (NOT a string) |
+| `enabled` | `boolean` | Yes | Must be `true` to connect |
+| `environment` | `object` | No | Environment variables |
+
+### Example with Environment Variables
+
+```json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "mcp": {
+    "neuronlayer": {
+      "type": "local",
+      "command": ["node", "C:\\path\\to\\neuronlayer\\dist\\index.js", "--project", "."],
+      "enabled": true,
+      "environment": {
+        "NEURONLAYER_MAX_TOKENS": "8000",
+        "NEURONLAYER_DEBUG": "true"
+      }
+    }
+  }
+}
+```
+
+---
+
+## Available Tools (51 Total)
+
+Once connected, OpenCode's AI can use all NeuronLayer tools:
+
+### Core Tools (5)
+| Tool | Description |
+|------|-------------|
+| `get_context` | Get relevant codebase context for a query |
+| `search_codebase` | Semantic search across all files |
+| `record_decision` | Save an architectural decision |
+| `get_file_context` | Get content of a specific file |
+| `get_project_summary` | Overview of project structure |
+
+### Symbol & Dependency Tools (3)
+| Tool | Description |
+|------|-------------|
+| `get_symbol` | Find functions, classes, types by name |
+| `get_dependencies` | See imports/exports for a file |
+| `get_file_summary` | Get compressed file summary (10x smaller) |
+
+### Learning Tools (3)
+| Tool | Description |
+|------|-------------|
+| `get_predicted_files` | Files predicted to be relevant |
+| `get_learning_stats` | Usage statistics |
+| `mark_context_useful` | Feedback for learning |
+
+### Multi-Project Tools (6)
+| Tool | Description |
+|------|-------------|
+| `list_projects` | List all registered projects |
+| `switch_project` | Switch active project |
+| `search_all_projects` | Search across all projects |
+| `discover_projects` | Find projects on system |
+| `record_decision_with_author` | Decision with author attribution |
+| `update_decision_status` | Update decision lifecycle |
+| `export_decisions_to_adr` | Export to ADR markdown files |
+
+### Feature Context Tools (4)
+| Tool | Description |
+|------|-------------|
+| `get_active_context` | Get current work session context |
+| `set_feature_context` | Start tracking a feature |
+| `list_recent_contexts` | List recent work sessions |
+| `switch_feature_context` | Switch to previous session |
+
+### Living Documentation Tools (7)
+| Tool | Description |
+|------|-------------|
+| `generate_docs` | Generate documentation |
+| `get_architecture` | Get project architecture overview |
+| `get_component_doc` | Get detailed component docs |
+| `get_changelog` | Get changelog of recent changes |
+| `validate_docs` | Check for outdated docs |
+| `what_happened` | Query recent project activity |
+| `find_undocumented` | Find code lacking docs |
+
+### Context Rot Prevention Tools (4)
+| Tool | Description |
+|------|-------------|
+| `get_context_health` | Check context health |
+| `trigger_compaction` | Manually compact context |
+| `mark_critical` | Mark content as critical |
+| `get_critical_context` | Get critical items |
+
+### Confidence Scoring Tools (3)
+| Tool | Description |
+|------|-------------|
+| `get_confidence` | Get confidence score for code |
+| `list_sources` | List sources for a suggestion |
+| `check_conflicts` | Check for decision conflicts |
+
+### Change Intelligence Tools (4)
+| Tool | Description |
+|------|-------------|
+| `what_changed` | Query what changed |
+| `why_broke` | Diagnose why something broke |
+| `find_similar_bugs` | Find similar bugs from history |
+| `suggest_fix` | Get fix suggestions |
+
+### Architecture Enforcement Tools (7)
+| Tool | Description |
+|------|-------------|
+| `validate_pattern` | Validate code against patterns |
+| `suggest_existing` | Find existing functions |
+| `learn_pattern` | Teach a new pattern |
+| `list_patterns` | List all patterns |
+| `get_pattern` | Get pattern details |
+| `add_pattern_example` | Add example to pattern |
+| `get_architecture_stats` | Get architecture statistics |
+
+### Test Awareness Tools (4)
+| Tool | Description |
+|------|-------------|
+| `get_related_tests` | Get tests for a file/function |
+| `check_tests` | Check if change breaks tests |
+| `suggest_test_update` | Get test update suggestions |
+| `get_test_coverage` | Get test coverage |
+
+---
+
+## Troubleshooting
+
+### "MCP server failed" or won't connect
+
+**Most common cause:** Relative paths don't work reliably.
+
+```json
+// ❌ WRONG - Relative path
+"command": ["node", "dist/index.js", "--project", "."]
+
+// ✅ CORRECT - Absolute path
+"command": ["node", "C:\\Users\\you\\neuronlayer\\dist\\index.js", "--project", "."]
+```
+
+### "Configuration is invalid"
+
+Ensure `command` is an array of strings:
+
+```json
+// ❌ WRONG - Single string
+"command": "node dist/index.js --project ."
+
+// ✅ CORRECT - Array of strings
+"command": ["node", "dist/index.js", "--project", "."]
+```
+
+### Server starts but immediately fails
+
+**Check if NeuronLayer runs manually first:**
+
+```bash
+node C:\path\to\neuronlayer\dist\index.js --project C:\path\to\your\project
+```
+
+If you see errors, the issue is with NeuronLayer itself, not OpenCode.
+
+**Common errors:**
+- "Cannot find module" - Run `npm run build` first
+- "ENOENT" - Path is wrong
+- Database errors - Delete `~/.neuronlayer/projects/` and retry
+
+### "No tools available" after connecting
+
+1. Verify connection with `/mcp` inside OpenCode
+2. Ensure `enabled: true` is set in config
+3. Restart OpenCode after config changes
+
+### Check Server Startup Manually
+
+Test if NeuronLayer starts correctly:
+
+```bash
+# Windows
+node C:\path\to\neuronlayer\dist\index.js --project C:\path\to\project
+
+# macOS/Linux
+node /path/to/neuronlayer/dist/index.js --project /path/to/project
+```
+
+You should see:
+```
+NeuronLayer starting...
+Project: /path/to/project
+Indexing started...
+Indexing complete: XX files indexed
+NeuronLayer MCP server started
+```
+
+### View Debug Logs
+
+Run OpenCode with debug logging:
+
+```bash
+opencode --log-level DEBUG
+```
+
+### Reset NeuronLayer Data
+
+If you have persistent issues, clear the data:
+
+```bash
+# Windows
+rmdir /s %USERPROFILE%\.neuronlayer
+
+# macOS/Linux
+rm -rf ~/.neuronlayer
+```
+
+---
+
+## Example Usage in OpenCode
+
+Once configured, try these prompts:
+
+```
+> Search for authentication code in this project
+
+> What architectural decisions have been made?
+
+> Show me the project summary
+
+> Find the function that handles user login
+
+> What changed yesterday?
+
+> Why did the tests break?
+
+> Check if this code follows our patterns: async function fetchData() { ... }
+
+> Get tests related to src/auth/login.ts
+
+> Record a decision: We chose SQLite because it requires no server setup
+```
+
+---
+
+## Data Storage
+
+NeuronLayer stores data in:
+
+```
+~/.neuronlayer/
+├── registry.json          # Project registry
+└── projects/
+    └── {project-name}-{hash}/
+        ├── neuronlayer.db  # SQLite database
+        └── tier1.json      # Working context
+```
+
+---
+
+## Links
+
+- [OpenCode Documentation](https://opencode.ai/docs/)
+- [MCP Protocol](https://modelcontextprotocol.io/)
+- [NeuronLayer API Reference](./API.md)
+- [NeuronLayer Architecture](./ARCHITECTURE.md)
+
+---
+
+*Last updated: February 2026*

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "neuronlayer",
-  "version": "0.2.7",
+  "version": "0.2.10",
   "description": "MCP server that gives AI assistants persistent understanding of your codebase",
   "type": "module",
   "main": "dist/index.js",
@@ -9,7 +9,8 @@
   },
   "files": [
     "dist/**",
-    "wasm/**"
+    "wasm/**",
+    "doc/**"
   ],
   "scripts": {
     "build": "node esbuild.config.js",