Clipboard manager for developers with secret redaction

The idea

A menu-bar app that keeps a searchable history of everything copied to the clipboard — text, code snippets, short file paths, image thumbnails — stored in a local SQLite database. Before anything is written to disk, it runs a fast pattern check (API keys, AWS credentials, JWTs, private key headers, .env-style KEY=value lines) and either redacts the match or skips storing that entry entirely, configurable per pattern.

Why build this

Developers copy-paste constantly: error messages, curl commands, config values, half-finished commit messages, one-off tokens pulled from a dashboard to test locally. Built-in OS clipboard history (Windows, macOS Sequoia) has no search past a handful of items and no awareness that a sk-live-... string just got persisted to a synced clipboard-history file that might sync to iCloud or a work laptop. Paid tools like Alfred's clipboard or Raycast handle history well but aren't built for the specific worry of "did I just leave a production secret sitting in plaintext history for 30 days." A narrow, local-first tool that's paranoid by default about secrets is a gap solo devs and security-conscious teams actually feel.

Stack sketch

• Menu-bar shell: Tauri (Rust backend, thin web UI) for a small binary and low idle memory versus Electron
• Storage: SQLite with FTS5 for full-text search over history, WAL mode for write-heavy clipboard polling
• Clipboard watch: native OS clipboard APIs via Tauri's clipboard plugin, polling at ~300ms with content-hash dedup to avoid storing repeats
• Secret detection: a bundled ruleset ported from Gitleaks/TruffleHog regex patterns, run synchronously before any write; user-editable allow/deny pattern list in a local TOML config
• Search UI: a global hotkey (e.g. Cmd+Shift+V) opens a fuzzy-search palette (fuse.js or a Rust equivalent like nucleo) over recent history
• Sync: none by default — explicitly local-only; if cross-device sync is ever added later it should be end-to-end encrypted and opt-in, not a v1 concern

Scope for v1

• Menu-bar app for macOS only (Windows/Linux later)
• Plaintext and image clipboard capture, no rich-text/RTF handling
• Secret detection covering the ~15 highest-signal patterns (cloud provider keys, private key blocks, generic high-entropy strings, .env assignments)
• Redact-in-place display (show sk-live-****) with a manual "reveal and copy anyway" override for false positives
• Retention cap: configurable max items (default 500) or max age (default 14 days), auto-pruned
• Global hotkey search + click-to-recopy
• No cloud sync, no team sharing, no browser extension companion

Where it could go

Once the local core is solid, the natural extension is a lightweight team mode: an opt-in local network or self-hosted relay so a team can share a scoped "snippets" clipboard (SQL queries, common commands) while keeping personal clipboard history fully separate and never leaving the machine. A second direction is IDE integration — a VS Code extension that reads from the same history store so pasting into a terminal or editor gets the same search palette without alt-tabbing to a separate app. There's also room for smarter classification beyond regex: a tiny local model that flags "this looks like a stack trace" or "this looks like a SQL migration" for better auto-tagging in search.

Watch out for

Secret-detection regexes have real false-positive and false-negative rates — a v1 that's too aggressive will redact legitimate hex strings and UUIDs and annoy users into disabling the feature entirely, while one that's too lax gives false confidence. Ship the pattern list as transparent and user-tunable from day one, and default to "redact and let the user reveal" rather than "silently drop," so nothing important vanishes without a trace.