wallfacer

wallfacer is an open-source terminal session manager built in Go by Pradipta, released under the MIT license. It indexes AI coding sessions from Claude Code, Cursor CLI, Kiro CLI, and Codex into a single unified view, letting you organize and navigate sessions you would otherwise lose track of across dozens of untitled transcript files. The project reached v1.2.3 as of August 2026 and is available via Homebrew or go install.

What It Is

wallfacer solves a concrete problem: every major AI coding agent stores its sessions in its own opaque directory structure — Claude Code uses JSONL files under ~/.claude/projects/, Cursor CLI buries chats in hash-named directories, Kiro CLI uses a flat folder, and Codex files rollouts by date. After a few weeks, you have dozens of transcripts you can't tell apart. wallfacer scans all four locations, reads just the head of each file for metadata, and keeps your own overlay of titles, tags, and projects in a local SQLite database — without ever modifying the agents' files.

How It Works

wallfacer operates as a read-only indexer of agent session directories. It stores user-defined metadata (titles, tags, project groupings) in its own SQLite database at ~/.local/share/wallfacer/, so deleting the database loses only the overlay, never a conversation. Sync is incremental for performance with hundreds of sessions. The rm command moves sessions to wallfacer's own trash; only --purge permanently deletes. For agents like Kiro CLI and Cursor CLI that spread one session across multiple files, the whole set travels together on delete.

Two Front Ends, One Index

wallfacer ships as a single binary with two modes:

• TUI (wallfacer with no arguments): a full-screen interactive browser with a session list on the left and a detail pane on the right. Supports fuzzy filtering, project/tag/agent cycling filters, inline rename/tag/project editing, and session launch/resume.
• CLI (wallfacer <command>): one-shot subcommands for scripting — new, resume, list, search, show, rename, tag, project, rm, and sync. All commands support --json output for piping.

Both modes read and write the same SQLite index, so changes made in the TUI are immediately visible to CLI commands and vice versa.

Multi-Agent Support

wallfacer currently supports four AI coding agents side by side:

• Claude Code — sessions under ~/.claude/projects/
• Cursor CLI — chats under ~/.cursor/chats/
• Kiro CLI — sessions under ~/.kiro/sessions/cli/
• Codex — rollouts under ~/.codex/sessions/

Agents are pluggable adapters, and the roadmap lists opencode as the next planned addition. The --agent filter accepts case-insensitive substrings, so filtering by kiro or cursor works without exact matching.

Update: v1.2.3

The latest release is v1.2.3, published on August 1, 2026. The project was created in late July 2026 and has seen rapid iteration. The roadmap includes full-text search across session content via SQLite FTS5, a wallfacer restore command for un-trashing sessions from the CLI, Markdown transcript export, per-project/week stats, and a longer-term desktop app built on Wails and xterm.js using the same Go internals. A daily update check queries GitHub for newer releases and surfaces the result in the TUI footer or on stderr for subcommands, with no blocking network calls; it can be disabled via WALLFACER_NO_UPDATE_CHECK=1.