The local-first email agent
Letterknife reads your email on your own computer, turns it into tasks, chases the follow-ups, and hands you a plan every morning. It runs on the Claude subscription you already have — and nothing goes out without your OK.
$1/month for your first 3 months, then $19 · Cancel anytime · It runs its builder's inbox every single day
Letterknife runs on your computer. Gmail stays the only place your email lives — nothing flows through our servers, so there is nothing of yours on them to leak, mine, or train on. Privacy by architecture, not by policy.
Hosted AI email tools bake their inference bill into the price. Letterknife orchestrates the Claude subscription you already pay for — so you get far more AI per dollar, and we charge for the software, not the tokens.
Labeling and archiving is the easy part. Letterknife tracks who owes you a reply, keeps every thread as a task with a next step, budgets your call time, and turns the pile into a plan you can actually run.
Not screenshots — the real interface, shown with sample data. Three steps from connected to autopilot.
Step 1 — Train
No setup interview, no forms. It shows you a real email, makes its call, and you correct it in plain words. Every correction becomes a rule it keeps — including the exceptions, which is where your real preferences live. You can watch it converge in one sitting.
New comment in Sailing Founders: "Great breakdown, one question about…"
Your correction
"My own community IS important — I'm the admin there. Everything else from Skool, just learn from."
Step 2 — Track
Letterknife doesn't stop at labeling. It keeps each conversation as a task, knows what's currently being asked of you, and has the reply drafted in your voice before you open it. Sending is consent-gated: it asks, you approve.
🌱 first raised Jun 12 · 🕑 last reply 2h ago · Petra Kovač
The ask
Petra sent redlines and is waiting on your answer to clause 7 (payment terms, net-30 → net-45).
Draft ready — in your voice
Hi Petra, net-45 works if we keep the early-payment discount at 2%. I've accepted the rest of your redlines. Updated draft attached — if that works, let's sign this week…
Step 3 — Run your day
Calls, replies, approvals and chases — merged with your calendar into one plan with a visible time budget. Things others are driving sit in a watch list instead of cluttering your day. Work the plan; the inbox takes care of itself.
We charge for the software, not the tokens. Hosted AI email tools bake their inference bill into $18–42/month. Letterknife runs on the Claude subscription you already pay for.
for your first 3 months, then $19/month — cancel anytime
Nowhere. Letterknife runs on your computer and talks to Gmail directly. Your mail is never uploaded to our servers — we could not read it if we wanted to. The agent’s own notes and rules stay on your machine too.
Your own Claude subscription. Letterknife orchestrates Claude on your machine — you don’t buy AI credits from us and we never mark up tokens. No Claude subscription? You can bring your own API key instead.
No. Sending is consent-gated: it drafts, you approve. As trust builds you can loosen the leash for specific senders or actions, but the default is always ask first.
Those are hosted services: your mail flows through their servers and their AI spend is baked into the subscription. Letterknife is the opposite architecture — the agent runs where your mail already is, on AI you already own. And it goes past triage: tasks, follow-ups, and a daily plan, not just labels.
Gmail and Google Workspace today — and you can connect as many Gmail accounts as you want under one plan. Your login is separate from the inboxes it manages. Outlook is on the roadmap.
$1/month for your first 3 months while we are in beta, then $19/month. The switch is automatic and visible on your invoice from day one — no surprises. Cancel anytime.
Your mail stays on your machine. Your AI does the work. You say yes.