dsh-shuttle

Developers who juggle DSH, Codex, Claude Code, Reasonix, Pi, and OpenCode on a daily basis know the pain: conversation history is locked into each platform's private format, so switching tools, moving machines, or simply archiving a thread means losing the context you built with that assistant. DSH Shuttle exists to break those silos. It projects every supported source into a provider-neutral intermediate model, then serialises the result either as DSH events or as the target platform's native import format, preserving text, reasoning traces, tool calls and results, timestamps, working directory, and provider/model identity wherever the destination can express them.

Migration is fully bidirectional: you can import external sessions into DSH in bulk, or export DSH tasks back to any supported destination. Safety is the default posture. Every operation is a dry-run preview first; a JSON report is produced for you to review, and nothing is written until you explicitly pass --apply. Existing DSH IDs and destination files are always skipped, never overwritten; reads are bounded by size and count and never recurse through symlinked directories; OpenCode's database is touched only through its official JSON export/import envelope, never edited directly. Three interaction paths are available out of the box: the visualised import/export forms in the Settings panel with independent previews, the /shuttle slash command rendered in the conversation UI without round-tripping to a model, and the offline CLI for scripted or CI-driven workflows.

If you work across two or three AI coding tools simultaneously and want a single, queryable archive in DSH's JSONL or SQLite backend, or if your team splits between Claude Code and Reasonix and you need to hand off debugging context without re-typing every prompt, DSH Shuttle is the bridge. It does not replace any tool in the chain; it simply makes sure your conversations can move safely and reversibly between them.