Public Roadmap

Built-in feedback submission action for Claude Threads so that users can submit feature requests and bugs. Hooks into APIs from Compass.

Addresses opportunity: "Prospective and new users cant self-serve answers about plugin capabilities" (Opportunity ID 77f18437-f031-4908-806f-f05e2a31ba4f). Reduces support load and improves evaluator-to-installer conversion; reinforces the Community Plugins submission.

We support both Claude and Codex now, and probably shouldn't have Claude in our name anyhow for trademark issues. We started this work but didn't finish it

Addresses opportunity: "New installs dont reach a working first multi-agent run" (Opportunity ID 5fd0998c-7fa4-42e3-98b5-6e08f4153aaa). Reduces install-to-abandon risk by getting users to a successful first agent run faster.

Addresses opportunity: "New users cant discover the plugin through Obsidian itself" (Opportunity ID b68ec554-6870-400d-a13c-316d879dbb7c). Getting listed in the Community Plugins browser is the primary top-of-funnel channel for Obsidian users.

Preserved candidate, not a delivery commitment. Distinct contextual documents would open as native tabs in one companion group; reopening a document would focus its existing tab. The first split would start near 30/70, the divider would be draggable, and user-adjusted sizing would persist. Validation is still required for Geode tab-group and split APIs, duplicate and accumulation behavior, and the default ratio.

LATER candidate, not committed. Users would be able to send the exact rendered screenshot, viewport, diagnostics, and source digest back to Claude or Codex for critique; preserve immutable iterations; compare them; and restore an accepted version without losing later history. Highest-priority future Design mode candidate because it closes the main usability gap found after Phase 1. Requires validation of whether exact visual feedback improves satisfactory outcomes and whether version controls remain lightweight.

LATER candidate, not committed. Users would be able to iterate on real React/Vite-style components or repository-backed designs through a supervised preview runtime with health checks, bounded ports, logs, teardown, security policy, accessibility summaries, and explicit dependency behavior. Validate demand and complete the visual-review loop before admitting this operationally complex path to NEXT.

LATER candidate, not committed. Users would be able to supply design tokens, reusable components, screenshots, and style constraints, then hand an accepted design into a production repository with traceability and less manual recreation. Optional reference-import or publish adapters remain explicit integrations rather than making Design mode a deployment platform. Validate handoff demand before ranking this candidate.

Shipped 2026-08-30. Users can place a singleton Claude Threads conversation in the main workspace while contextual vault files open in a native companion leaf. The release preserves activeThreadId and reload state, restores Classic placement safely, protects edited files, and was verified in packaged Geode at small and maximized window sizes. Release: https://github.com/rbcodelabs/obsidian-claude-threads/releases/tag/v0.29.0. Implementation: https://github.com/rbcodelabs/obsidian-claude-threads/pull/457.

Shipped 2026-08-21 in Claude Threads v0.27.4. Users can enter /design with a brief, create a durable local HTML/CSS/JS artifact, preview it in Geode’s sandboxed ArtifactView, see live reload and diagnostics, switch responsive viewports, capture a PNG, and reopen the artifact from its thread. Release: https://github.com/rbcodelabs/obsidian-claude-threads/releases/tag/v0.27.4. Implementation evidence: PR #413 (workflow) and PR #414 (release). This is the Phase 1 foundation; visual review/versioning remains a separate LATER candidate.

Addresses opportunity: "Users have no visibility into whats coming or how to influence it" (Opportunity ID 124201da-f65b-4a3d-96a0-35ad66d7157b). Builds trust and gives the community a low-friction prioritization channel.