Switch frontier coding agents in the same chat, without losing context.
Cadrega brings your coding agents into one workspace, preserving context, memory, and progress as you move between them.
Provider switching without losing the thread
Cadrega does not just switch models: it switches the provider working on the project. You can start with Codex, hand off to Claude Code, then continue with Grok or another runtime without breaking the session. The point is not "choose a different model"; it is keeping the same line of work: messages, touched files, decisions, tools, approvals, and context stay readable in one record.
Cadrega-history: resuming a conversation for real
When you resume a session, Cadrega should not pretend to remember. It checks the record first. cadrega-history lets the provider read the current conversation, inspect recent steps, search keywords, and reconstruct what was decided. It is memory as an audit trail: precise, inspectable, and tied to the session.
Vector Memory: search for meaning, not just words
After a long session, keyword search is not enough. Vector Memory indexes the chat and enables broader, faster semantic search: you can find "the handoff decision", "the context problem", or "the provider section" even when those exact words are not in the record. Cadrega uses vector memory first to orient itself, then cadrega-history to confirm exact details.
PRIVATE BETA
Join the first cohort.
Cadrega is opening to a small group of developers and teams already using multiple coding agents in the same workflow. Get early access, test real agent handoffs, and help shape the product.
Frequently Asked Questions
Same project. Same memory. Same thread.
Switch agents without restarting the work.
Switch agents without restarting the work.