MetaLoom has two independently configured runtime components. This page is a hub; follow the links for the full settings tables.
| Component | How it is configured | Reference |
|---|---|---|
Loom (server) |
|
|
Cortex (worker daemon) |
Environment variables / CLI flags (optional |
Loom
Loom loads loom.yml from the first location that exists — /etc/metaloom/loom.yml,
~/.config/metaloom/loom.yml, then config/loom.yml — and then applies environment-variable
overrides on top, so any LOOM_* variable wins over the file. Configuration groups:
-
database — PostgreSQL connection + pool (
LOOM_DB_*) -
server — REST/gRPC/MCP/monitoring ports and bind address (
LOOM_SERVER_*) -
auth — keystore, token lifetime, OAuth2, MCP auth (
LOOM_INITIAL_PASSWORD,LOOM_TOKEN_EXPIRATION_TIME,LOOM_OAUTH2_*,LOOM_MCP_AUTH_*) -
storage — upload directory for asset binaries (
LOOM_STORAGE_UPLOAD_DIR) -
ai — the Chat & AI agent (
LOOM_AI_*) -
sandbox — the agentic coding sandbox (
LOOM_AGENT_SANDBOX_*) -
memory — the agent memory bank (
LOOM_AGENT_MEMORY_*)
See Loom Configuration for the full YAML example and every environment variable with defaults. The AI-agent, coding-sandbox and memory settings are also covered on the Chat & AI Agent page (which also covers the coding sandbox).
Cortex
A Cortex worker is configured with environment variables (or the equivalent CLI flags): LOOM_HOST /
LOOM_PORT (default 8092) to reach Loom, LOOM_TOKEN to authenticate, CORTEX_MONITORING_PORT,
CORTEX_META_PATH, CORTEX_NODE_ID, and CORTEX_NODE_WHITELIST / CORTEX_NODE_BLACKLIST to control
which node kinds it runs. See Cortex Configuration.
Precedence
For both components, environment variables override file/default values. This makes container
deployment straightforward: set the handful of LOOM_* / CORTEX_* variables you need and skip the
config file entirely.