// the find
msitarzewski/agency-agents-app
Agency Agents is a small, native app for browsing, installing, and tracking the agent personas from msitarzewski/agency-agents across the AI coding tools you actually use.
A Tauri 2 + Svelte 5 desktop app that manages installation of AI agent persona files (markdown/TOML system prompts) across coding tools like Claude Code, Cursor, Copilot, and others. It solves a real coordination problem: each tool expects its agent configs in a different location and format, and this app keeps a local ledger so you know what's installed, what's drifted, and what needs updating. For developers who juggle multiple AI coding tools and want consistent agent personas across them without manually copying files.
The reconciliation model is genuinely well-thought-out — it hashes both the source render and the installed file, so it catches drift from manual edits rather than just checking file existence. The Rust backend handles all file writes and IPC, and the README is explicit that the frontend cannot construct arbitrary shell commands, which is the right call for a Tauri app with filesystem access. Tool knowledge lives in a single `tools.json` shared between backend and frontend, so adding a new tool doesn't require touching both layers independently. The offline-first approach with a bundled corpus baseline means it works without network access after install, with optional sync.
The app is fundamentally a distribution mechanism for one specific upstream catalog (`msitarzewski/agency-agents`), so its usefulness is entirely coupled to that repo's quality and maintenance trajectory — if that catalog goes stale or the author loses interest, this app becomes a polished installer for outdated configs. Windows support exists but isn't code-signed yet, which means SmartScreen warnings on every install; that's a meaningful friction point for a tool targeting developers who presumably have security-conscious installs. Project-scope installs are only available for Cursor and opencode as of v0.2.0, while several tools that are arguably more popular for project-level agent config (Copilot, Claude Code) are global-only. The memory-bank design docs committed to the repo suggest heavy Claude Code usage as an authoring tool, which is fine, but it also means the repo's architecture docs are essentially LLM session notes — not a substitute for proper architectural documentation.