Your project, your codebase, is the whole context for your AI.
Manage that context well. Lantern will help you with that.
Lantern maps your project, helps you plan changes, reviews pull requests with generated flows and local review, and keeps your AI assistant sessions across your harness together in one desktop app.
- Watches every session across your harness and shows a live dashboard of each one: its project, state, and last reply.
- Notifies you in the menu bar and with native notifications when a session finishes, needs approval, or is waiting for input, so you don't babysit a window.
- Maps your project into a
.lantern/folder (one record per file, its role, its purpose and its relations) and draws it as a graph, so you can see the structure without reading the code. - Reviews pull requests and local changes with a file list, diffs and the graph, in review order, with a reason for each changed file.
- Local review: review your uncommitted changes, or a branch with no pull request, with the
same file list, diffs and graph as a pull request (+ Compare in the Lantern tab, or
lantern --compare). One click asks Claude for a one-sentence reason per changed file. - Generate with Claude or Codex: choose the assistant used by Lantern's maps and repairs for each project. This setting does not change the assistant inside the embedded editor. Pick the output language per project and repair broken output in place.
Sessions are told apart by their working directory: one entry per VS Code project window.
Packages are attached to each release on the
releases page: Lantern-mac.zip for
Apple-silicon Macs and Lantern-win-x64.zip for Windows. The Apple-silicon build is also published
as Lantern-mac-arm64.zip for the updater, and the Lantern-mac.zip link stays the one to use. Until a release is published there,
build Lantern from source (installation).
The macOS app of a release built by the current workflow is signed with a Developer ID certificate and notarized by Apple, so macOS only asks once to confirm opening it. Earlier releases are signed ad hoc and need the steps in CONTRIBUTING.md. The Windows build is not code-signed, so SmartScreen may warn on first launch.
The documentation is at https://docs.thelantern.dev and in docs/:
- Installation and command line
- Data and privacy: what Lantern stores, how to remove it, and what leaves your machine
- Getting started
- Architecture
- Product overview: thelantern.dev
The checks CI runs on every push and pull request (.github/workflows/ci.yml, a plain Linux
runner with Node 22, pnpm 10.28.0 and pnpm install --frozen-lockfile) are:
pnpm run typecheck
pnpm run lint # biome lint --diagnostic-level=error
pnpm testCI runs the unit tests, builds on Linux, macOS and Windows, runs the Playwright e2e for integrations, the tray, pull request host state and skills (the graph suite is not run in CI), and publishes a coverage report as a build artifact. It signs nothing. The stack, the project layout, the code and testing conventions and the build pipeline are in CONTRIBUTING.md; a map of the source is in docs/architecture.md.
- CONTRIBUTING.md: how to work on the app, test it and build it.
- SECURITY.md: how to report a vulnerability.
- CODE_OF_CONDUCT.md: how we expect people to behave in this project.
- TROUBLESHOOTING.md: known problems and their fixes.
- THIRD_PARTY_NOTICES.md: licences of the software Lantern ships.
Lantern is released under the MIT License. Copyright (c) 2026 Mobile Reality. See LICENSE. Contact: hello@themobilereality.com.
Lantern is not affiliated with or endorsed by Anthropic, OpenAI, Microsoft, Atlassian or GitHub; their names and marks are trademarks of their owners.
Made with <love> by People with AI
By Mobile Reality