Owlat is modular — every feature listed in this guide can be turned on or off. This page is the user-facing overview of how to do it.

Feature flags

Owlat ships as a single codebase but you choose which features run on your install. Marketing campaigns, the team inbox, AI assistance, Postbox personal mail — every one of those is a feature you opt into, not a product you have to commit to up front.

Some flags imply infrastructure

The sending flags (campaigns, transactional, automations) only work with a configured delivery provider (MTA, Resend, or SES) — a connected external mailbox is not one. See Operating Modes for the coherent combinations and the rules Owlat enforces.

This page is the operator's reference. For the underlying mechanics (single source of truth, dependency resolution, docker profiles) see Feature flags — developer reference.

Where to toggle

There are three places to flip a flag: the admin UI, the host CLI, and the first-run setup wizard.

Admin UI

Settings → Features in the dashboard. Each flag has a description, dependency hints, and a toggle. Disabling a flag with dependants pops a confirmation listing what will also turn off.

Setup CLI

On a VPS install the host wrapper (scripts/owlat, installed as owlat) exposes the flag commands:

# Turn one flag on or off:
owlat feature postbox on
owlat feature ai off

# Set a required env var, then apply:
owlat env LLM_API_KEY sk-...
owlat restart

These run the containerized TS CLI (owlat-setup) inside the wolvesdotink/setup image. Each toggle rewrites the CLI-side flag mirror (.owlat-flags.json), regenerates docker-compose.override.yml so the right background workers are selected, and tells you to run owlat restart to apply.

The inner CLI also has a pack command for toggling several flags at once — it is only available on owlat-setup directly, not via the owlat wrapper:

owlat-setup pack marketing on
owlat-setup pack emailClient on

To inspect the current state, read .owlat-flags.json in your install directory or open Settings → Features in the dashboard. There is no --show/list command.

Setup wizard

The first-run web wizard (served by the wolvesdotink/setup container, reached via owlat setup or the install one-liner) has a "Pick what to enable" step (the first step, after the welcome screen) that presents the three feature packs plus every individual flag grouped by category. Toggle a pack or flip flags one at a time; dependency cascades apply as you go. You can always change the answers later with the admin UI or the CLI above.

The older bash scripts/setup.sh wizard (owlat setup --legacy) does not configure feature flags — it only handles deployment mode, env vars, and the email provider.

Feature packs

A pack is just a button that toggles several related flags at once. Three packs ship with Owlat:

Pack	Flags	Use when
Marketing	`campaigns`, `automations`, `transactional`	You want to send marketing emails and trigger-based flows.
Email Client	`inbox`, `chat`, `postbox`	You want a shared team mailbox, chat surface, and personal mail (Postbox).
AI	`ai`, `ai.agent`, `ai.autonomy`, `ai.knowledge`, `ai.knowledge.autoLink`, `ai.knowledge.graphRetrieval`, `ai.knowledge.analytics`, `ai.assistant`, `ai.visualizations`	You want every AI feature on at once.

Toggling a pack walks each member flag through the normal dependency cascade, so you can't get into an invalid state.

Dependencies and cascades

Some flags require others. The rules:

Turning on a dependent flag auto-turns-on its prerequisites. Enabling ai.agent also enables ai and inbox.
Turning off a flag turns off everything that depended on it. Disabling ai cascades off ai.agent, ai.autonomy, ai.knowledge (and its ai.knowledge.autoLink, ai.knowledge.graphRetrieval, ai.knowledge.analytics sub-flags), ai.assistant, and ai.visualizations.

Flag	Requires	Notes
`campaigns.archive`	`campaigns`	Archive links need campaigns
`inbox.codeTasks`	`inbox`, `ai.agent`	Needs the inbox and the AI agent
`ai.agent`	`ai`, `inbox`	Agent classifies + drafts for inbox mail
`ai.autonomy`	`ai`, `ai.agent`	Autonomy extends the agent
`ai.knowledge`	`ai`
`ai.visualizations`	`ai`

The bare `ai` flag also powers Postbox in-inbox AI

Postbox's Summarize thread and Suggest replies (Reading mail) gate on the top-level ai flag (not ai.agent). Turn ai off and those buttons stop working — the actions are also rate-limited per user. As with all AI, you need an LLM provider configured (LLM_API_KEY).

What turning a flag on changes

Toggling a flag has three concrete effects:

The UI hides or shows sidebar nav, settings panels, and onboarding tasks for the feature.
Background workers start or stop — but only on the wizard and CLI paths. Each flag can declare Docker Compose profiles. Examples: postbox → personal-mail; mail.external → external-mail; scan.files → clamav; inbox.codeTasks → inbox-codetasks; ai → ai. The setup wizard and the owlat feature/owlat-setup pack commands recompute the active profile set and regenerate docker-compose.override.yml, so an owlat restart actually starts or stops the worker. Toggling a flag in the admin UI only persists the flag in apps/api (instanceSettings.featureFlags); it does not regenerate the override or restart containers. To bring a worker up or down from the dashboard, re-run the relevant owlat feature command (or the wizard) on the host and owlat restart.
Required env vars are surfaced. Some flags need extra environment variables (e.g. ai requires LLM_PROVIDER and LLM_API_KEY; scan.urls requires GOOGLE_SAFE_BROWSING_API_KEY; analytics.posthog requires POSTHOG_API_KEY and POSTHOG_HOST). The setup wizard prompts for them. The admin UI can't read your .env, so it does not block the toggle — it shows an advisory modal reminding you which vars to set with owlat env <KEY> <VALUE> (then owlat restart).

All flags

Flag	Category	Default	What it enables
`campaigns`	Sending	On	Marketing campaigns to contacts and segments
`campaigns.archive`	Sending	On	Public "View in browser" links
`transactional`	Sending	On	Programmatic sends via the API
`automations`	Sending	Off	Multi-step trigger workflows
`inbox`	Receiving	Off	Shared team inbox with threading
`inbox.codeTasks`	Receiving	Off	Bug-report extraction from inbound mail
`chat`	Receiving	Off	Real-time chat alongside inbox
`postbox`	Receiving	Off	Personal mailboxes with IMAP/SMTP
`mail.external`	Receiving	Off	Connect a user's own external Gmail/Fastmail/company mailbox over IMAP+SMTP (no sending domain required)
`ai`	AI	Off	Master toggle (needs LLM provider)
`ai.agent`	AI	Off	Classify + draft replies
`ai.autonomy`	AI	Off	Agent sends without human approval
`ai.knowledge`	AI	Off	Semantic knowledge graph
`ai.knowledge.autoLink`	AI	Off	LLM-inferred typed edges between entries (on top of the always-on rule-based linker); requires `ai.knowledge`
`ai.knowledge.graphRetrieval`	AI	Off	Expand search results along graph edges before grounding drafts (kill switch); requires `ai.knowledge`
`ai.knowledge.analytics`	AI	Off	Centrality/cluster cron + knowledge-graph dashboard; requires `ai.knowledge`
`ai.assistant`	AI	Off	Multi-turn streaming tool-calling AI assistant + @assistant replies in team chat
`ai.visualizations`	AI	Off	NL-prompt-to-chart dashboards
`webhooks`	Integrations	Off	Outbound event delivery
`forms`	Integrations	On	Embeddable signup/capture forms
`imports.mailchimp`	Integrations	Off	One-click Mailchimp import
`imports.stripe`	Integrations	Off	Stripe customer sync
`scan.content`	Security	On	Spam / phishing / homoglyph pre-send check
`scan.files`	Security	On	ClamAV attachment malware scan
`scan.urls`	Security	Off	Google Safe Browsing URL reputation
`analytics.posthog`	Deliverability	Off	PostHog product analytics
`domains.verification`	Deliverability	On	SPF/DKIM/DMARC validation gate
`domains.dkimRotation`	Deliverability	On	Scheduled DKIM key rotation

Some flagged behaviors aren't fully live yet

A few flags gate features whose surface is shipped but whose deeper automation isn't wired end to end:

ai.autonomy — the agent config (auto-reply, confidence threshold, tone, signature, daily cap) is real, but graduated autonomy rules and the autonomy feedback loop have no production callers yet. See AI Agent & Autonomy.
inbox.codeTasks — code tasks can be created and managed manually, but auto-creation of a task from an inbound email is not wired. See Code Tasks.

When in doubt

You can always turn a flag on, look around, decide it's not for you, and turn it back off — there's no migration penalty for either direction. Data attached to a feature (mailboxes, campaigns, automations) is preserved when you toggle a flag off; turning it back on restores access.

Hosted-cloud flags (billing.stripe, multiTenancy, tier.autoProvision) are hidden in self-host mode and don't appear in the admin UI.