Elephant is a context-preservation layer for AI coding agents that lets you switch between tools like Claude Code and Codex without losing the work already done. Instead of manually explaining a long conversation, repo structure, bug investigation, code changes, failed tests, and current objective every time an agent hits its usage limit or context window, Elephant continuously records the session while the agent is working. When you open the same repository with another agent, Elephant restores the important working state—what the previous agent was trying to achieve, what it discovered, which files it changed, what remains broken, and where to continue—so the new agent can pick up almost exactly where the previous one stopped. The outgoing model does not need to generate a handoff summary; Elephant has already captured the context while it was alive.
Features
- Context Recovery — Restores previous agent working context.
- Agent Switching — Move between Claude and Codex.
- Session Recording — Captures context while agents work.
- Zero Handoff — No manual summaries between agents.
- Repo Awareness — Remembers files, changes, and structure.
- Objective Tracking — Preserves the current development goal.
- Bug Memory — Retains discovered bugs and reasoning.
- Change Tracking — Records files modified during sessions.
- Test State — Remembers passing and failing tests.
- progress Recovery — Continues from the exact stopping point.
- Cross-Agent Context — Shares state across different coding agents.
- Automatic Continuity — New agents immediately understand progress.