Social repliesLive

Larkreply

Answer every reply to your posts from one list, in your own voice, without handing your accounts to a bot.

Role
Founder. Product, design, front-end, back-end, browser extension, desktop apps and deployment.
Team
Solo build. I am user number one.
Timeline
Designed and built 18 to 19 November 2026
Stack
Node.js (no framework, server-rendered HTML), Supabase (Postgres, Auth, Storage), a SQLite to Postgres translation layer (node:sqlite on the desktop), Vercel functions and cron, Electron with electron-updater, a Chrome MV3 extension, Expo and React Native, WebCodecs and mp4-muxer, Playwright, Claude API
Year
2026
6
platforms in one inbox: LinkedIn, X, Threads, Bluesky, Mastodon and Substack
Scope fact
4
surfaces: web, Mac, Windows and Linux desktop, a Chrome extension, and mobile in progress
Scope fact
199
automated tests, run against both SQLite and Postgres
Delivery fact
0
replies sent without your click. On LinkedIn, X, Threads and Substack you press Send yourself.
Design rule

The problem

Schedulers made posting easy. Replies did not get the same help. They land in six different apps, in six different notification feeds, and the ones that matter get buried under likes.

The replies are where trust gets built, so they are the half that should not be skipped. I wanted one list: every reply twice a day, the post it is about, a draft in my voice, and me deciding what goes out.

Key decisions

Constraint

LinkedIn and X suspend accounts that post through automation, and a reply tool is only useful if your account is safe.

Options

Automate sending everywhere, skip the platforms without a reply API, or keep a person on the Send button.

Decision

Bluesky and Mastodon send through their official APIs. On LinkedIn, X, Threads and Substack a Chrome extension reads your notifications in your own signed-in browser, and Copy & open puts the approved reply in the comment box. You press Send.

Trade-off

One more click per reply on four platforms. In return, nothing is ever posted by a bot and no account is put at risk.

Constraint

People want a desktop app that keeps data on their own computer, and a web app they can open anywhere.

Options

Two codebases, or one codebase that runs on two databases.

Decision

One Node.js codebase with a small translation layer between SQLite and Postgres. The desktop app runs on node:sqlite; the web app runs the same code on Supabase. The test suite runs against both.

Trade-off

Every query has to work in two dialects. In return, a fix ships to desktop and web at the same time.

Constraint

Studio makes animated brand cards for every post, and rendering video on a server costs money per minute.

Options

Pay a rendering service, run headless browsers in the cloud, or render on the user’s machine.

Decision

Cards render in the browser with WebCodecs and mp4-muxer and download as MP4s that open on the finished card, so a muted autoplay still reads.

Trade-off

A browser without WebCodecs falls back to a slower real-time recording. In return, no rendering bill and no render queue.

Constraint

A new account needs to show what the product does before anything is connected, but demo content must never be mistaken for real replies.

Options

An empty state, a video tour, or labelled demo data.

Decision

Load demo replies fills the inbox with clearly labelled demo items, a DEMO CONTENT ribbon on each one, and they disappear on their own once a real account connects. Sending a demo reply only marks it sent.

Trade-off

More states to design and test. A new user sees a full inbox in one click and never confuses it with their own.

Constraint

People paste their own Claude key and platform tokens into it.

Options

Store secrets as entered, or treat every secret and every outside request as a risk.

Decision

Per-user keys are encrypted before they are stored, every query is scoped to the signed-in user, and the import-a-brand-kit-from-your-website feature is guarded against requests to private network addresses.

Trade-off

More code around small features. Each person’s key is only ever used for their own drafts.

The work

Larkreply inbox with replies from six platforms, a reply selected, the original post expanded and a draft in your voice
InboxEvery reply from six platforms in one list, the post each one is about, and a draft in your voice to approve, edit or skip. Copy & open puts it in the comment box and you press Send. Demo data
Larkreply Queue week grid with scheduled posts, open slots and a draft ready to schedule
QueueA week grid of time slots. Drag a draft onto an open slot, and posts go out at their time through each platform’s official posting API. Demo data
Larkreply brand kits with two card previews, colour pickers with contrast checks and a gradient picker
Brand kitsColours with a contrast check, gradients, 30 Google fonts, 20 label styles and 16 layouts for the animated cards. A kit can be imported from your website. Demo data
Larkreply analytics showing which posts started the most conversations
AnalyticsWhich posts start the most conversations, across every platform they went to, and how many replies got an answer. Demo data
Larkreply Studio library of posts with their platforms and Queue status
StudioPaste notes, an article or a link and Claude writes a version for each platform, inside each platform’s limits. Demo data
Larkreply settings with the voice guide Claude reads before drafting
Voice guideClaude reads it before every draft, written the way you would brief a sharp assistant. Demo data
Larkreply help with a five-step quick start and a screenshot guide for creating a Claude API key
HelpFive steps from sign-up to the first reply, with a screenshot guide for each one. Demo data

How it was built

  1. 01DiscoverCalls and walk-throughs with the people who would use it
  2. 02CompileTranscripts, files and constraints in one requirements brief
  3. 03SpecifyA detailed, sourced spec from Perplexity, reviewed and locked
  4. 04BuildDesigned and built with the latest AI models, in the production stack
  5. 05TestUser testing on every preview, with real users on real tasks
  6. 06IterateFindings turned into the next round of changes, worst problems first
  7. 07ShipTests, security review and release, then back to testing and iterating↻ Back to Test
How I work, step by step →

Where AI was used

Claude drafts every reply and every platform version with the person’s own API key, guided by a voice guide they write. It never sends anything: a person approves every reply and presses Send. I designed and built it, working with AI coding agents I directed, reviewed and tested.

What’s next

The iPhone, iPad and Android app (Expo) for answering replies on the go, then Studio video rendering in the web app.