FAQ

What omoikane does, what it doesn't, and how the privacy posture actually works in practice. Direct answers to the questions people actually ask.

What does omoikane do?

Helps career-switchers, market re-entrants, and neurodivergent job seekers run a careful, evidence-cited application workflow:

1. Match gate — score your fit against a posting before drafting anything. <50% means stop and save your energy.
2. Tailored CV — drafted from your own uploaded evidence. Every claim cites the source.
3. Cover letter — written in your voice, anchored to your samples — never AI-default.
4. Application email — sent from your address, with optional reply tracking.

Plus post-application support: contact tracker, calendar, neurodivergence-aware scheduling, optional onboarding 30/60/90 once you’re hired, end-to-end encrypted private diary.

What does omoikane NOT do?

• Not a job board. We don’t list openings. You bring the postings.
• Not a recruiter tool. Recruiters don’t have access to your data; we don’t sell to them.
• Not a CV-template generator. Templates exist, but the value is the per-posting tailoring.
• Not a general AI assistant. Each agent is scoped to one step in the workflow.
• Not a “growth-hacked” SaaS. No nudging emails, no engagement metrics, no in-app upsells.

Why is BYOM-Free the default?

BYOM = “Bring Your Own Model”. When you sign up, the default tier is BYOM-Free: you bring your own LLM API key (from one of seven supported providers — see pricing) and we route your requests through your account so the cost lands on your bill (typically $0–$5/month for active job-search use).

We chose this default because charging unemployed users to recoup our $20–60/month of LLM compute is structurally broken. BYOM keeps the product accessible without burning operator runway. If you’d rather we manage the LLM cost instead, the Managed-Lite tier (€5–9/mo) covers $5/mo of compute on our keys.

What’s a “tracker alias”?

When you send an application email through omoikane, we offer to CC a per-application alias address (app-<token>@omoikane.coach) so replies from the recruiter land back in your application-tracker view inside omoikane instead of getting buried in your personal inbox.

Opt-in per application — default ON, with a toggle on the email-coach preview to turn it off per draft. Disable globally in Settings → Integrations → Reply tracker.

We do not use inbound replies as training data, do not forward them to third parties, do not auto-reply on your behalf without your explicit click-through.

Why Gmail OAuth (and not just SMTP)?

When you connect Gmail at signup, we ask only for gmail.send + gmail.compose scopes. That lets us draft + send via your real Gmail, so the recipient sees an email from you (not from a bot relay). Reply-All from the recipient hits both Gmail (your inbox) and the optional CC tracker alias.

Voice samples can also be drag-imported from Gmail via incremental-auth: the gmail.readonly scope is only requested at the moment you click “Import from Gmail” and revoked when the dialog closes. We never silently request read access at signup.

Outlook works equivalently via Microsoft Graph.

How does the diary stay private?

The diary feature uses per-user envelope encryption for body text at rest:

• Each account gets a unique data-encryption key (DEK) generated at signup.
• The DEK is wrapped (encrypted) with our operator-held key-encryption key (KEK), which is sops-encrypted at rest in our infrastructure secrets bundle.
• The wrapped DEK is stored alongside your user record; the plaintext DEK never persists.

Operator access to your diary requires going through an audited route that writes an entry to the operator_actions audit log before any decryption. You can see this audit log from your account settings.

Even an attacker with a database snapshot cannot read diary bodies without also obtaining the operator KEK from sops; even an operator with KEK access cannot read a diary entry without leaving an audit-log trace you can see.

What’s the data-residency story per LLM provider?

When you supply a BYOM key, your CV / posting / coaching / draft data is sent to your chosen provider per the provider’s own privacy policy (not omoikane’s). At provider-add time we surface a clear data-residency banner so you can make an informed choice:

For our Managed tiers, we route through Anthropic Claude (United States data residency). Choose Mistral via BYOM if EU-only data residency is your hard requirement.

Can omoikane really write in MY voice?

That’s the whole pitch. The letter-coach is grounded on your voice samples: anything you upload (cover letters, blog posts, sample emails, posts) at /profile/voice. The drafted letter should read like you wrote it on a good day.

With zero voice samples, the letter-coach falls back to AI-default voice and you’ll see a banner saying so — that’s the soft-gate from Sprint G1. We block letter drafting entirely until you’ve completed the coaching profile (4 short exercises, ~10–15 minutes) and uploaded at least one voice sample.

What’s the “soft gate” that won’t let me draft a letter?

Two prerequisites must be satisfied for the letter-coach (and email-coach) to run:

1. Coaching profile complete — 4 short exercises (values, motivators, qualities, STAR-R stories). Captures who you are and how you communicate.
2. At least one voice sample on file — pasted text or uploaded file (.txt, .docx, .pdf, .rtf, .eml, .md).

Without these, the system cannot draft authentic content for you. Match gate + CV draft still work without the soft-gate prerequisites — only the voice-tier outputs (letter + email) are gated.

What happens if my BYOM key fails?

Structured error returned, banner shown in the workspace, email digest if persistent. We never silently fall back to our operator-paid LLM keys when a BYOM user’s key fails — that would surprise-charge our account on a tier where you are meant to pay.

If your key hits its monthly cap (/profile/ai-budget), the same hard-block applies: no LLM call dispatched, structured error shown.

What does the safety chain do?

Every AI output goes through a 5-stage chain: persona-shift detection (catches AI-default voice), citation enforcement (claims must cite uploaded evidence), output scrubber (per-language banned phrases), canary detection (HTML-comment indirect-injection traps), PII scanning (phone / IBAN / email / IP).

The chain is surface-aware post-2026-05-09: cover letters and emails are user voice, so we don’t flag normal hedge phrases like “In summary” or “Please feel free to reach out”. CV claims and gate reasoning are factual, so citation enforcement applies strictly there. Diary uses voice-tier rules.

If something gets flagged, we surface it to you with the specific signal — never silently rewrite your work.

What languages does omoikane support?

UI: English, Dutch, Greek. Pick yours in Settings.

The AI agents (gate / cv / letter / email) run in your selected language end-to-end — including the safety chain (per-language persona-shift patterns + scrubber tables for EN / NL / EL).

How does omoikane handle calendar integrations?

Three providers supported: Google Calendar, Microsoft Outlook Calendar, Nextcloud (CalDAV). Per-user multi-select — you can have 0, 1, 2, or 3 connected.

Push-only by default — we write events (deadline reminders, follow-up reminders, interview slots) to your calendar(s) but don’t read your busy slots. Two-way is opt-in per integration — when enabled, omoikane reads busy slots and uses them to suggest interview times.

What’s the Onboarding 30/60/90 thing?

When you mark a posting as outcome=hired, omoikane offers to activate Onboarding mode: a 90-day series of operator-curated Daily Focus tasks designed to help you land well in the new role. Three phases (first-30 days, 60-day, 90-day) with time-suggested + user-confirmed transitions.

Optional. Decline at the activation prompt and it never bothers you again.

What’s “Chill Mode”?

When your energy_state (set in the ND Support panel) is tired or burnt_out for 3 consecutive entries, omoikane proactively switches to a low-pressure interface: greys out non-urgent tasks, suggests a one-week pause, surfaces self-saved searches as a passive feed instead of a to-do list.

Designed for the realistic energy curves of long job searches, especially for neurodivergent users.

What’s “ND Support”?

A panel that tracks your energy_state across 5 levels (energetic → good → tired → burnt_out → crisis) plus a per-language nd_tips corpus tied to specific job-search triggers (interview anxiety, rejection emails, decision fatigue, etc.).

In crisis state, the entire UI pauses and surfaces NL mental-health resources (113.nl, MIND, Sensoor) inline. We don’t push you to use omoikane when you shouldn’t be using anything.

How does cancellation / data deletion work?

From your account settings:

• Export everything in a zip, anytime — 18-table export including drafts, voice samples, calendar events, contacts, diary entries (decrypted with your DEK).
• Delete everything, anytime — irreversible, cascades through every backing system within 7 days.

We log every operator access to your data; you can see the operator-actions log from settings.

Where can I read more?

• Privacy policy — full GDPR-compliant detail with sub-processors register
• Security page — what we explicitly do NOT do, sub-processor list with country and transfer-basis for each
• How it works — the 4-step workflow walkthrough
• Pricing — tier comparison

If a question isn’t answered here, write to hello@omoikane.coach.

← Back to home