Keep your coding context alive.
ContinueVault carries architecture decisions, debugging history, repo conventions, failed approaches, and constraints across your AI sessions — chat, web, or CLI.
ContinueVault is not a code indexer. It does not scan your repo, watch your IDE, or upload your source tree. It saves the AI conversation layer you choose to preserve, so future AI sessions can continue the work.
Re-explaining your codebase, stack, and conventions to every new chat or CLI session.
Debugging sessions that hit dead ends and disappear with no trace.
Architecture decisions whose reasoning you remember vaguely but can't reconstruct.
Switching from Claude to ChatGPT — or from a chat to a CLI — and rebuilding the entire technical brief.
Supported continuation workflows can use what was tried and ruled out.
Prior debugging paths and their results stay searchable.
"Why did we pick X" has an answer instead of a guess.
Code reviews can continue in any supported AI workflow without losing nuance.
What works today
ContinueVault syncs selected supported web AI accounts and captures the conversation layer from supported CLI tools. You can bring that context back into ChatGPT through the ContinueVault GPT, or into any ContinueVault MCP-connected AI tools. See supported sources and continuation tools below for the current list.
By default, ContinueVault captures the conversation around your code from CLI sessions — prompts, AI reasoning, decisions, dead-ends — but does not capture file contents, diffs, tool inputs, tool outputs, or command results. Full session sync — including tool inputs, tool outputs, file contents, diffs, and command results — is available only as an explicit opt-in for users who need it.
A codebase is a different trust class than an AI conversation. ContinueVault's default is to preserve continuity without turning itself into a code indexer, repo scanner, IDE monitor, or terminal recorder.
You spent forty minutes in Claude Code last Tuesday narrowing a race condition to a specific event listener — ruled out the queue ordering theory, hypothesized stale cache invalidation, decided on a transactional fix.
Today the symptom resurfaces and you want a second opinion from ChatGPT.
ContinueVault gives ChatGPT — through the ContinueVault GPT — the prior errors, the failed fixes, the constraints, and your current hypothesis. ChatGPT picks up where Claude Code left off. Your actual code never left git.
ContinueVault captures the conversation around your code — the decisions, the dead-ends, the reasoning that doesn't show up in commit messages. Your IDE, your terminal, and your code itself stay where they belong: in git. By default, even when we capture from CLI tools, we don't capture file contents or command outputs.
- Continue Thread — bring relevant prior context into ChatGPT through the ContinueVault GPT, or any ContinueVault MCP-connected AI tools.
- Threads across platforms — group related debugging and design sessions captured from any supported web AI or CLI tool.
- Unified search — find prior decisions, debugging hypotheses, rejected approaches, and implementation constraints across saved AI conversations.
- Code stays out by default — tool inputs, tool outputs, file contents, diffs, command results, and source code are excluded from CLI capture unless full session sync is explicitly enabled.
- All-time historical sync/backfill (Pro) — sync full available history from connected web AI accounts, extract decisions and constraints, and make them searchable. CLI capture follows your local cv-sync configuration.