Import your full X/Twitter data archive, generate semantic embeddings, search by meaning — then batch delete with surgical precision. All from your terminal, all on your machine.
Download your data from X, then let x-molt turn it into a searchable, cleanable local database.
Drop in your X data export (.zip or .js files). Auto-discovers structure, validates with Zod, loads into local SQLite.
Run Transformers.js locally to create vector embeddings for every tweet, DM, like, and note. Zero API calls.
FTS5 for exact keywords. Semantic search for concepts like “regrettable hot takes”. Browse by date tree.
Filter by date, keywords, replies, retweets. Preview, then delete at ~180/hr with pause, resume, and ETA.
Archive features are completely free — you only pay X’s API cost when you delete.
Handles .zip and extracted .js from X’s data export. Auto-discovers file structure, validates with Zod schemas, stores tweets, DMs, likes, followers, and ad data in local SQLite.
Generate vector embeddings locally via Transformers.js. Runs on your CPU — no API calls, no data leaving your machine. Embeds tweets, DMs, likes, and notes.
FTS5 full-text search for exact keyword matches. sqlite-vec semantic search for concepts and meaning. Date tree navigation with split-pane detail views.
Filter by date range, keywords, replies, retweets. Preview before committing. Rate-limited to ~180/hr for safety. Pause/resume with ETA. Idempotent — tracks deletions.
Activity heatmaps, engagement patterns, privacy audit, and ad targeting analysis. Multiple tabbed dashboards surfacing insights from your archive data.
Every API call shows its price upfront. Session, daily, monthly, and all-time dashboards with per-endpoint breakdowns and sparkline charts.
Beyond the archive pipeline, x-molt is a complete X terminal client with transparent pay-per-use pricing.
Browse home feed and own tweets with inline actions.
Compose with 280-char counter, preview, post.
Search recent tweets across all of X.
Split-pane conversations, compose, send.
Browse, follow, unfollow, block, mute.
View, create, and manage your lists.
Browse and manage bookmarked tweets.
Like, retweet, bookmark from any view.
~8,500 lines of strict TypeScript. Clean layers, modular services, full type safety.
No third-party services. No cloud. No telemetry. Just a TypeScript TUI running locally.
Your entire archive, embeddings, and cost history live in a single database on your filesystem.
Vector embeddings generated locally via Transformers.js. Your tweet content never hits an external API.
Credentials stored with chmod 600 in ~/.config/x-molt/. Direct OAuth 1.0a with X — no middleman.
Every API call shows its cost upfront. Cost badges on every menu item. Full spend dashboards.
From install to semantic search in four steps.
Run npx x-molt or npm i -g x-molt. Requires Node 18+.
Go to X → Settings → Your Account → Download an archive.
Point x-molt at your .zip. It parses, validates, and stores everything. Then generate embeddings for semantic search.
Search by keyword or meaning. Browse by date. Batch delete what you don’t want.