Configuration
Updated at 2026-09-21 · 623d849
User-specific paths and settings pass through documented environment
variables — no package hardcodes a machine path or guesses where data
lives. This table is rendered from the registry in
packages/shared/contracts/src/env.ts on every release.
Most installs only ever set one: DIGITAL_ME_WIKI_ROOT.
| Variable | Required | Default | Description |
|---|---|---|---|
DIGITAL_ME_HOME | no | derived from DIGITAL_ME_WIKI_ROOT when unset (legacy alias) | Legacy alias for DIGITAL_ME_WIKI_ROOT (the data root: wiki + config). Optional — when unset it falls back to DIGITAL_ME_WIKI_ROOT. Prefer exporting DIGITAL_ME_WIKI_ROOT. |
DIGITAL_ME_WIKI_DIR | no | resolved relative to DIGITAL_ME_HOME / DIGITAL_ME_WIKI_ROOT at load time | Directory containing your wiki .md files. |
DIGITAL_ME_WIKI_ROOT | no | ~/digital-me | Canonical wiki/data root that every runtime adapter, hook, dream-cycle, and dashboard-intake reads (contains wiki/, inbox/, config.yaml). This is the variable users export; DIGITAL_ME_HOME is the contracts-layer alias. |
DREAM_CYCLE_HOME | no | — | Directory for the dream-cycle Python pipeline. |
DREAM_CYCLE_VENV | no | — | Python venv used by dream-cycle. |
OPENCLAW_HOME | no | ~/.openclaw | Openclaw configuration and data home. |
OPENCLAW_DATA_DIR | no | — | Openclaw runtime data directory (SQLite stores). |
OPENCLAW_GATEWAY_HOST | no | 127.0.0.1 | Host of the openclaw gateway HTTP server. Defaults to the IPv4 loopback literal (not 'localhost', which Node >=17 may resolve to ::1 and miss a 127.0.0.1-bound gateway). |
OPENCLAW_GATEWAY_PORT | no | 18789 | Port of the openclaw gateway HTTP server. |
OPENCLAW_GATEWAY_TOKEN | no | — | Auth token for the openclaw gateway. If unset, read from $OPENCLAW_HOME/openclaw.json. |
DIGITAL_ME_BRAIN_URL | no | — | digital-me brain-host /tools/invoke endpoint (e.g. http://127.0.0.1:18791/tools/invoke). When set, every brain caller uses it instead of the openclaw gateway; the bearer token must resolve from DIGITAL_ME_BRAIN_TOKEN, DIGITAL_ME_BRAIN_TOKEN_FILE or the default token file (a URL with no token is a hard error, never a gateway fallback). |
DIGITAL_ME_BRAIN_TOKEN | no | — | Bearer token for DIGITAL_ME_BRAIN_URL. Prefer DIGITAL_ME_BRAIN_TOKEN_FILE in service units and MCP registrations so the secret stays on disk (mode 600). |
DIGITAL_ME_BRAIN_TOKEN_FILE | no | — | File holding the bearer token for DIGITAL_ME_BRAIN_URL, read when DIGITAL_ME_BRAIN_TOKEN is unset. Default: <DIGITAL_ME_WIKI_ROOT>/.data/brain-host.token (written by digital-me install --runtime brain-host). |
DIGITAL_ME_BRAIN_DB | no | resolved by resolveBrainDbPath(): /.data/brain.db, else the legacy <OPENCLAW_HOME>/data/brain.db while only that exists | Path to brain.db (goals, tasks, traces, learnings). When unset, every reader applies the same rule: <DIGITAL_ME_WIKI_ROOT>/.data/brain.db if it exists, else the legacy <OPENCLAW_HOME>/data/brain.db if it exists, else the canonical <DIGITAL_ME_WIKI_ROOT>/.data/brain.db. |
DIGITAL_ME_ENV_FILE | no | resolved by resolveEnvFilePath(): /.data/.env, else the legacy <OPENCLAW_HOME>/.env while only that exists | Env file holding provider keys (GEMINI_API_KEY, …), loaded by the brain-host service and inherited by the workers it dispatches. When unset: <DIGITAL_ME_WIKI_ROOT>/.data/.env if it exists, else the legacy <OPENCLAW_HOME>/.env if it exists, else <DIGITAL_ME_WIKI_ROOT>/.data/.env. |
BRAIN_PROXY_PATH | no | — | Absolute path to the brain-mcp-proxy binary. If unset, resolved via PATH. |
ORCHESTRATOR_DB_PATH | no | — | Path to brain-orchestrator SQLite store. |
DASHBOARD_PORT | no | 3458 | Port the dashboard Express server listens on (loopback only). |
DASHBOARD_TITLE | no | Operations Dashboard | Title shown in dashboard UI header. |
TEAM_WORKSPACE_ROOT | no | — | Optional root directory of a shared team workspace. Dashboard team views are disabled if unset. |
LEARNING_SOURCE_DIR | no | — | Optional source directory for learning artifacts to ingest into the dashboard. |
LEARNING_DEST_DIR | no | — | Optional destination directory for processed learning notes. |
OPENCLAW_AGENT_ID | no | unknown | Caller agent identity for brain attribution. Set per-runtime in MCP server env block. |
Workflow dispatch to your CLIs is configured separately in
~/digital-me/config.yaml under cli_exec_aliases — digital-me setup
pre-populates it for detected CLIs.