Home · Quick start · Editor · charter.yaml · Sources · Agent · Guides · Evals · Audit · Policies · CLI · MCP · Workspace · Desktop · About · FAQ
DataCharter as a real application: a native window over the same local server —
no terminal, no Python, no browser tab. Pick a workspace folder (or start with
the demo), and everything works exactly as it does under datacharter serve:
the explorer, guides, evals, audit, the governed agent surface.
The Windows build, captured automatically on every release by CI.
The app opens with an 11-step guided tour covering the whole arc: query your data, see what an agent sees, guides, policies, the audit chain, and evals.
Download (beta)
Grab the artifact for your platform from the latest release:
| Platform | File |
|---|---|
| macOS (Apple Silicon) | DataCharter-<version>-macos-arm64.dmg |
| Windows 10/11 | DataCharter-<version>-windows-x64.zip |
Intel Macs: use brew install datacharter/tap/datacharter or uvx datacharter —
GitHub retired free Intel macOS runners, so we don’t ship an Intel build.
The app remembers your last workspace and keeps a Workspace ▸ Recents menu.
brew install datacharter/tap/datacharter and uvx datacharter remain the
primary, friction-free paths — the app is for the days you don’t want a terminal.
macOS: opening an unsigned beta
These builds are not yet signed with an Apple Developer ID, and macOS Sequoia removed the old right-click→Open shortcut. First launch:
- Double-click the app; macOS will refuse with “Apple could not verify…”. Close it.
- Open System Settings → Privacy & Security, scroll to the Security section.
- You’ll see “DataCharter.app was blocked…” — click Open Anyway, then confirm Open in the dialog.
That’s a one-time step per version. (Signing/notarization is on the roadmap; Homebrew and uvx installs have no such friction meanwhile.)
Windows note
The .exe is unsigned, so SmartScreen may warn: click More info → Run
anyway. The Windows build is verified by automated smoke checks in CI but has
not yet had long human soak time — treat it as beta and report anything odd.
How it works
The app is the same DataCharter you install from PyPI — a local FastAPI server plus the web UI — wrapped in your OS’s native webview (WKWebView on macOS, WebView2 on Windows). It binds to localhost only, and every governance layer (masking, row filters, read-only guard, audit, canaries) applies identically.