Developer robinovitch61 released Jeeves, a terminal UI that lets users browse, preview, and resume AI agent sessions without leaving the command line. It aggregates conversations from Claude and Codex into a single list, letting engineers quickly locate prior prompts or code suggestions. By surfacing history in a split‑pane view, the tool reduces context‑switching during development.

Jeeves supports regex‑based search, and direct session resumption via a simple keystroke. Users can filter by keywords such as “ui refactor” or pattern “fix.*bug” and instantly open the full transcript. The interface respects Claude’s default 30‑day cleanup policy, but developers can extend retention by editing ~/.claude/settings.json, ensuring long‑term auditability.

Installation spans Homebrew, Nix, Arch AUR, Go, and Windows package managers, giving users multiple entry points. A simple `brew install robinovitch61/tap/jeeves` drops a binary into the PATH, while `go install github.com/robinovitch61/jeeves@latest` compiles from source. Open‑source under the MIT license, Jeeves invites contributions for additional AI frameworks.

By consolidating Claude and Codex dialogues, Jeeves addresses a common pain point: scattered AI outputs across separate CLI tools. Teams can now audit model behavior, reuse snippets, and maintain a reproducible development trail, all from a familiar terminal environment. The project’s modular design promises future adapters for Anthropic or OpenAI agents.