About Reinstate
What Is Reinstate? Coding-Agent Work Continuity
Reinstate is an open-source continuity layer for coding-agent work. Stable v0.2.0 moves encrypted Claude Code and Codex sessions between configured devices and adds configless local search plus same-vendor native resume/fork. Stable v0.3.0 adds verified resume. Stable v0.4.0 adds explicit structured cross-agent handoff after dual-platform tagged-artifact acceptance PASS. Stable v0.5.1 adds universal agent coverage by support tier after dual-platform tagged-artifact acceptance PASS. Stable v0.6.0 adds Reinstate Hop (cloud continuity) and the interactive CLI, certified by native Windows x64 tagged-artifact acceptance PASS under a single-platform waiver (ADR 0005); Apple Silicon macOS acceptance is deferred to issue #403.
- Last verified
- Current release
- v0.6.0
- Release status
- v0.6.0 stable · native Windows x64 tagged-artifact acceptance PASS (215/215 required rows, docs/testing/results/2026-09-09-windows-v060rc8.md) under the single-platform waiver (ADR 0005) · Apple Silicon macOS acceptance deferred to #403 · supersedes v0.5.1
- Owner
- Harjot Singh Rana
The problem Reinstate solves
Git moves source code, but it does not move the local history of a long coding-agent task: the rejected approaches, decisions, prompts, file context, and native session identifier. That history becomes fragmented across agents, projects, worktrees, operating systems, and devices.
Reinstate owns continuity before and after execution. Coding agents continue to own the actual execution loop. The goal is to let a developer find and continue a task in the right native environment without re-explaining it from scratch.
What is available today
Phase 1 implements the trust-sensitive sync engine and CLI surface for supported Claude Code and Codex CLI sessions:
- Same-vendor session discovery, push, pull, status, diff, and conflict handling.
- Local age encryption before upload to user-owned S3-compatible storage.
- Canonical project identity and macOS ↔ Windows structural path remapping.
- Credential exclusions, OS-keyring storage, restore backups, and atomic writes.
- Checksum-verifying installers and compatibility checks that fail closed.
- Configless local indexing, literal search, bounded inspect, and a TTY switcher.
- Native same-vendor Claude Code/Codex resume and fork; read-only discovery across a wider agent catalog by support tier.
- Phase 3 verified-resume reports and gates for repository, agent, capability, and runtime state.
- Phase 4 structured handoff into a new Claude Code or Codex session.
- Phase 5 universal agent coverage: one catalog, one support-tier ladder (T0–T5), honest T0.
Stable evidence stays platform-specific.
Stable v0.6.0 is physically verified on native Windows x64 (tagged-artifact acceptance PASS) under a single-platform waiver (ADR 0005); Apple Silicon macOS acceptance is deferred to issue #403 until that hardware returns. Earlier stable releases (v0.5.1 and before) passed dual-platform tagged-artifact acceptance on both Apple Silicon macOS and native Windows x64. Intel macOS and Linux/WSL2 packages remain preview and unverified rather than inferred support.
- Latest public source tag
- v0.6.0(stable; cloud continuity and the interactive CLI, certified native Windows x64 only)
- Latest stable release
- v0.6.0
- Initial public prerelease
- Supported operating systems
- Native Windows x64 verified for v0.6.0 (Apple Silicon macOS deferred, issue #403); both remain the supported platform boundary. Intel macOS and Linux/WSL2 preview
Where the product goes next
| Phase | Direction | Status |
|---|---|---|
| 1 | Encrypted same-vendor Claude Code and Codex session sync | Stable since v0.1.0 |
| 2 | Local universal session index, search, inspect, resume, and last-session commands | Stable v0.2.0 |
| 3 | Verified resume against repo, branch, runtime, MCP, and skill state | Phase 3 stable; dual-platform verified resume |
| 4 | Explicit portable cross-agent checkpoints and capability diffs | Stable since v0.4.0 |
| 5 | Universal agent coverage: one catalog, one support-tier ladder, honest T0 | Stable since v0.5.0/v0.5.1 |
| 6C | Cloud continuity: Reinstate Hop (sign-in, locker, device pairing, daemon) | Stable v0.6.0; native Windows x64 tagged-artifact acceptance PASS, Apple Silicon macOS deferred (#403) |
| 6A/6B | Universal non-secret configuration and automatic cross-device reconciliation | Planned for a later release |
| 7 | Project continuity: durable rules and decisions rendered to every agent, plus learned memory with provenance | Planned for v0.8.0, after Phase 6A/6B |
| 8+ | Optional thin console, then team continuity | Exploring |
Later roadmap work is not present-day functionality. In particular, Reinstate does not provide cross-agent transcript translation, environment repair, configuration sync, or team handoffs.
What Reinstate is not
- Not a code editor, terminal emulator, or full agentic development environment.
- Not a multi-agent scheduler, model router, or worktree orchestrator.
- Not a Reinstate-owned plugin runtime or marketplace.
- Not raw mirroring of entire vendor configuration or credential trees.
- Not silent Claude Code ↔ Codex transcript conversion.
Reinstate should integrate with the places developers already code. Its product boundary is whether work survives, can be found, and can continue safely—not where that work executes.
Who Reinstate is for
The primary audience is developers using terminal coding agents across multiple sessions, projects, or environments. A work desktop and personal laptop is the flagship multi-device case, especially across Windows and macOS, but physical device count is not the entire product identity.
The long-term north-star metric is previously started coding tasks successfully resumed per active user—not storage volume or device count.
Project and maintainer
Reinstate was created and is maintained byHarjot Singh Rana. The project is available onGitHub under the Apache License 2.0 and welcomes review, issues, and contributions.
Reinstate is independent of Anthropic, OpenAI, Google, xAI, and other agent vendors. Vendor names describe compatibility targets, not sponsorship or endorsement.