Get Started
Configuration
AgenticReality is configured through environment variables. No configuration file is required. The engine auto-detects the deployment environment and applies sensible defaults.
AgenticReality is configured through environment variables. No configuration file is required. The engine auto-detects the deployment environment and applies sensible defaults.
Core Environment Variables
| Variable | Default | Description |
|---|---|---|
AREAL_STORAGE | ~/.agentic/reality/ | Directory for .areal state files |
AREAL_MCP_TOOL_SURFACE | compact | MCP tool surface: compact or full |
AGENTIC_TOKEN | (none) | Bearer token for HTTP/SSE authenticated mode |
AGENTIC_REALITY_VERSION | (latest) | Pinned version for install script |
AREAL_LOG | warn | Log level: error, warn, info, debug, trace |
AREAL_SENSE_INTERVAL_MS | 30000 | Auto-sense interval in milliseconds |
Deployment Environment Detection
AgenticReality auto-detects the deployment environment from standard signals. You can override with AREAL_ENV:
| Variable | Values | Description |
|---|---|---|
AREAL_ENV | production, staging, development, ci, sandbox, simulation | Force a specific environment tier |
NODE_ENV | production, staging, development | Detected automatically |
CI | true | Detected automatically (GitHub Actions, GitLab CI, CircleCI) |
SANDBOX | true | Detected automatically |
SIMULATED | true | Detected automatically |
MCP Server Configuration
The MCP server reads from standard MCP client config files. Add to your client config:
Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"agentic-reality": {
"command": "areal-mcp",
"args": []
}
}
}Claude Code (~/.claude/settings.json):
{
"mcpServers": {
"agentic-reality": {
"command": "areal-mcp",
"args": []
}
}
}VS Code / Cursor (.vscode/settings.json):
{
"mcp.servers": {
"agentic-reality": {
"command": "areal-mcp",
"args": ["--storage", "~/.agentic/reality"]
}
}
}Compact vs Full Tool Surface
The AREAL_MCP_TOOL_SURFACE variable controls how many MCP tools are exposed:
| Value | Tool Count | Use Case |
|---|---|---|
compact (default) | 15 | Standard usage; one facade per domain |
full | 78 | Developer usage; all granular operations exposed |
Compact mode is strongly recommended. It reduces LLM confusion and aligns with the MCP consolidation contract.
Storage Paths
| Path | Contents |
|---|---|
$AREAL_STORAGE/state.areal | Current reality state |
$AREAL_STORAGE/history/ | Past incarnation snapshots |
$AREAL_STORAGE/anchors/ | Persisted reality anchors |
$AREAL_STORAGE/logs/ | Operation logs (if enabled) |
Logging
Set AREAL_LOG to control verbosity:
AREAL_LOG=debug areal workspace senseIn MCP server mode, logs go to stderr (not stdout) so they don't corrupt the JSON-RPC stream.
Authentication (HTTP/SSE mode)
For remote deployment with HTTP or SSE transport:
# Generate a token
export AGENTIC_TOKEN=$(openssl rand -hex 32)
# Start MCP server with auth
areal-mcp serve --mode http --port 3010 --auth
# Clients must send: Authorization: Bearer <token>