Roadmap and Status

Origin uses three practical status buckets: released, on main, and planned. Released means it appears in CHANGELOG.md or the current README install path. On main means merged but not necessarily released. Planned means a product direction, not a promise.

This distinction matters because the repository moves quickly. The website should help users understand what they can rely on today without hiding where the project is going.

The released public shape is local-first AI work memory: Claude Code plugin, MCP server, CLI setup, daemon, local memory, hybrid retrieval, spaces, Markdown artifacts, source-backed pages, git versioning, and cross-platform runtime support.

That is enough for the core loop: brief, capture, recall, handoff, distill, and inspect local artifacts.

Recent merged PRs show a strong focus on retrieval quality and memory maintenance. Most of that work is opt-in while it is evaluated.

The useful way to read the direction is not as a pile of toggles, but as one goal: make the next AI session receive the smallest, most relevant, most trustworthy context bundle.

The product docs should stay practical. Setup, daily workflow, capture quality, memory types, architecture, commands, CLI/service management, updates, upgrade notes, packages, platform support, HTTP API, API examples, spaces, graph context, source-backed pages, import and portability, local git history, models and keys, retrieval status, experimental flags, data and privacy, backup and migration, configuration, environment variables, diagnostics, FAQ, evaluation, desktop status, releases, testing, development conventions, and troubleshooting are the current core path.

The remaining gap is mature retrieval documentation once opt-in experiments become stable defaults.

For users, progress should show up as fewer cold starts, fewer repeated explanations, better recall, and more inspectable context. For maintainers, progress should show up in eval snapshots, smaller PRs, and clearer release notes.

The project avoids a broad workflow-suite shape. The goal is a focused local memory layer for serious AI work, not a general productivity operating system.