Configuration¶

Complete reference for environment variables, runtime settings, and configuration files.

Environment Variables¶

Core¶

Variable	Default	Description
`COGNIBRAIN_RUNTIME_ROOT`	`.cognibrain/`	Root directory for runtime state
`COGNIBRAIN_PORT`	`8787`	HTTP API port
`COGNIBRAIN_HOST`	`127.0.0.1`	HTTP API bind address

Authentication¶

Variable	Description
`MEMORY_API_KEY`	API key for daemon authentication
`COGNIBRAIN_API_KEY`	Alternative API key variable
`COGNIBRAIN_API_TOKEN`	Alternative token variable
`MEMORY_BEARER_TOKEN`	Bearer token for JWT/OIDC auth
`MEMORY_OIDC_ISSUER`	JWT/OIDC issuer URL
`MEMORY_OIDC_AUDIENCE`	JWT/OIDC audience

Storage¶

Variable	Description
`MEMORY_DB_URL`	Database connection string
`MEMORY_POSTGRES_URL`	PostgreSQL connection string

Policy¶

Variable	Default	Description
`MEMORY_POLICY_MODE`	`development`	`production` for default-deny policy

Connectors¶

Variable	Description
`MEMORY_SLACK_TOKEN`	Slack bot token
`MEMORY_GITHUB_TOKEN`	GitHub personal access token
`MEMORY_VENDOR_LIVE_SMOKE`	Enable live connector smoke tests
`MEMORY_VENDOR_LIVE_WRITE`	Enable write operations in live smoke

Runtime State¶

Runtime data is stored under the runtime root (default: .cognibrain/):

.cognibrain/
├── config.json          # Runtime configuration
├── memories.json        # Local memory store (local-json mode)
├── connectors/          # Connector state
├── service/             # Service management state
└── logs/                # Local logs

Profiles¶

Profiles pre-configure a set of defaults for common deployment shapes:

`solo-dev` (default)¶

storage: local-json
auth: local-only
harnesses: [codex, cursor]
connectors: [github]
adapters: [storage-sqlite]
dashboard: off

`team`¶

storage: sqlite
auth: api-key
harnesses: [codex, cursor, vscode, neovim]
connectors: [github, slack, jira]
adapters: [storage-postgres, intelligence]
dashboard: off

`enterprise`¶

storage: postgres
auth: oidc
harnesses: all
connectors: [github, slack, jira, confluence, pagerduty]
adapters: [storage-postgres, intelligence, audit]
dashboard: optional

`benchmark`¶

storage: local-json
auth: local-only
harnesses: all
connectors: [github]
adapters: [benchmark]
dashboard: off

Harness Configuration Files¶

Generated by cognibrain init and cognibrain config:

Codex¶

File	Purpose
`$CODEX_HOME/config.toml`	Codex MCP and tool configuration
`AGENTS.md`	Agent instructions file

Cursor¶

File	Purpose
`.cursor/mcp.json`	MCP server configuration
`.cursor/rules/open-memory.mdc`	Cursor rules for memory integration

Package Harness¶

File	Purpose
`.cognibrain-harness-package.json`	Package-level harness metadata

Configuration Precedence¶

Configuration is resolved in this order (highest wins):

CLI flags (--port, --api-key, etc.)
Environment variables
.cognibrain/config.json
Profile defaults
Built-in defaults

Custom Configuration¶

Override settings in .cognibrain/config.json:

{
  "port": 9000,
  "storage": "sqlite",
  "auth": {
    "method": "api-key",
    "keys": ["env:MEMORY_API_KEY"]
  },
  "dreamCycle": {
    "autoRun": true,
    "frequencySessions": 10,
    "stalenessThresholdDays": 30
  },
  "contextDefaults": {
    "tokenBudget": 1200,
    "maxMemories": 20
  }
}

Security Configuration¶

For production deployments:

# Required
export MEMORY_API_KEY="your-secure-key"
export MEMORY_POLICY_MODE="production"

# Recommended for team/enterprise
export MEMORY_OIDC_ISSUER="https://auth.example.com"
export MEMORY_OIDC_AUDIENCE="cognibrain"
export MEMORY_POSTGRES_URL="postgresql://user:pass@host:5432/cognibrain"

Secrets

Never commit secrets to source control. Use environment variables, a secret manager, or the runtime-local connector config file.