Compatibility

Which AI coding tools Chronicle supports, what each feature does per tool, and where each tool's logs live on disk.

Chronicle imports conversation logs from six tools and maps every message to the Git snapshot at that moment. Most features work identically across all six; a handful of control-plane features (MCP Hub, Skills Hub) apply only to tools that keep the relevant config, and remote access is not built yet. Everything below reflects what ships in v0.1.7 — read the source key against server/parsers/<tool>.js if you need to go deeper.

Feature support matrix

Feature	Claude Code	Codex	Cursor	OpenCode	Gemini CLI	Copilot Chat
Conversation import	✅	✅	✅	✅	✅	✅
Time Travel / code snapshots	✅	✅	✅	✅	✅	✅
Replay Mode	✅	✅	✅	✅	✅	✅
Message filtering	✅	✅	✅	✅	✅	✅
Content redaction	✅	✅	✅	✅	✅	✅
Tool call viewing	✅	✅	✅	✅	⚠️	✅
Context Causality	✅	✅	✅	✅	✅	✅
Git history matching	✅	✅	✅	✅	✅	✅
Live streaming	✅	✅	✅	✅	✅	✅
MCP Hub takeover	✅	✅	✅	–	✅	–
Skills Hub takeover	✅	✅	✅	–	✅	–
Remote SSH access	🔜	🔜	🔜	🔜	🔜	🔜

Legend: ✅ full · ⚠️ partial · 🔜 planned (not yet built) · – not applicable.

• Tool call viewing is partial for Gemini CLI — its logs record tool activity less completely than the JSONL-based tools.
• MCP / Skills Hub takeover applies to the tools that keep a discoverable MCP/skills config that Chronicle can scan and centralize: Claude Code, Codex, Cursor, and Gemini CLI. OpenCode and Copilot Chat have no such config to take over.
• Remote SSH access (import / browse / live-watch over SSH) is planned but not implemented for any tool. Everything Chronicle does today runs against local files.

Log locations

Each parser reads its tool's native logs from a well-known path. Chronicle never writes to these — see the read-only column.

Tool	Source key	Path	Format	Read-only handling
Claude Code	claude-code	~/.claude/projects/	JSONL	Read in place; originals never modified
Codex	codex	~/.codex/sessions/	JSONL	Read in place; originals never modified
Cursor	cursor	VS Code workspaceStorage state DBs (CHRONICLE_CURSOR_DIR override)	SQLite (WAL)	Copied to temp including -wal/-shm before opening
OpenCode	opencode	~/.local/share/opencode/opencode.db	SQLite (WAL)	Copied to temp including -wal/-shm before opening
Gemini CLI	gemini-cli	~/.gemini/tmp/	JSON	Read in place; originals never modified
Copilot Chat	copilot-chat	VS Code workspaceStorage/<hash>/chatSessions/ (CHRONICLE_VSCODE_DIR override)	JSON	Read in place; originals never modified

Read-only, always: SQLite-backed sources (Cursor, OpenCode) are WAL databases. Copying only the .db file yields an empty database — recent writes live in the -wal sidecar — so the parsers copy the -wal and -shm files to a temp location and open the copy. Your tools' live databases are never touched.

Default path constants live in each parser (CLAUDE_PROJECTS_DIR, CODEX_SESSIONS_DIR, OPENCODE_DB, GEMINI_TMP in server/parsers/*.js). Only Cursor and Copilot expose environment overrides — see Configuration.

Per-tool caveats

• Gemini CLI records no working directory. Because there is no cwd in the logs, Chronicle assigns a virtual path (gemini-project:<hash>) and shows a "Needs association" banner. Point it at the real project once and Chronicle merges the sessions on path match.
• Copilot Chat spans VS Code distributions. The scanner looks in the workspaceStorage of VS Code stable, Insiders, and VSCodium, so Copilot sessions from any of those installs are picked up.
• Cursor and OpenCode share one database across sessions. Because one file backs many sessions, per-session source-file deletion is disabled for these tools (it is offered only for one-file-per-session sources: Claude Code, Codex, Copilot).

Known limitations

• Large sessions degrade gracefully. Beyond ~5,000 messages, the UI switches to windowed rendering — it draws roughly 400 DOM rows around your current position and decimates timeline ticks — so a 6,000-message session stays responsive. You navigate by search and the timeline rather than an unbounded scroll.
• Git submodules are supported by the snapshot engine.
• Non-standard or custom log paths are handled through manual selection: use the import wizard's Browse option (or the CHRONICLE_CURSOR_DIR / CHRONICLE_VSCODE_DIR overrides) to point Chronicle at logs outside the default locations.

Related

• Importing sessions — the import wizard and the read-only guarantees per source.
• Parsers & ingestion — the normalized event model and how to add a seventh source.