v0.5.2

What's Changed

• Keep the whole transcript visible and scrollable in the run view by @av in #41
• Release tracequest 0.5.2 by @av in #44

Full Changelog: v0.5.1...v0.5.2

v0.5.1

What's Changed

• Give the run view one way into analytics and a way back out by @av in #39
• Show Kandev-run fx and DeepSeek Harness sessions on their prompt, and live by @av in #42
• Release tracequest 0.5.1 by @av in #43

Full Changelog: v0.5.0...v0.5.1

v0.5.0

What's Changed

• Show usage collection errors in Plan limits by @av in #24
• Hide server-only Markdown and Export links in standalone reports by @av in #22
• Show finished Grok replies as idle instead of generating by @av in #29
• Explain empty chapter filters and let toolbar menus switch on click by @av in #31
• Count failed OpenCode tool calls so errors:>0 finds them by @av in #35
• Fix successful inspection being labelled as retries and waste by @av in #28
• Fix git activity option and commit message labels by @av in #26
• Preserve source and time filters when selecting an Insights machine by @av in #23
• Fix Grok detail and Insights failure-count disagreement by @av in #25
• Use one session grade in Runs, chat, detail and reports by @av in #27
• Fix Codex Send and Stop completing in one action by @av in #33
• Show the served session summary in CLI renders (Cursor duration and cost) by @av in #32
• Keep earlier chat messages and tool output reachable in finished conversations by @av in #34
• Open the failed call from What went wrong and name its cause by @av in #30
• Support vktr sessions through the grok code paths by @av in #36
• feat(web): make the run view read like Cursor's agent chat, with tool calls collapsed by default by @av in #37
• Support fx and DeepSeek Harness sessions by @av in #38
• Release tracequest 0.5.0 by @av in #40

Full Changelog: v0.4.0...v0.5.0

v0.4.0

What's Changed

• Fix duplicated tokens and incorrect model cost estimates by @av in #20
• Fix missing session messages, Codex tool order and cold loading stalls by @av in #19
• Remove colored left accents and flatten app corners by @av in #18
• Release tracequest 0.4.0 by @av in #21

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

• CI: run spec facts in parallel; skip timing asserts on the macOS release leg by @av in #14
• Keep only end-to-end facts and checks; run spec facts inline again by @av in #15
• docs: redo README screenshots with the 0.2.0 design by @av in #16
• tracequest 0.3.0: fast, stable hub on real data by @av in #17

Full Changelog: v0.2.0...v0.3.0

v0.2.0

What's Changed

• docs: show recording workflows with annotated screens by @av in #1
• ui + docs: cursor.com/brand restyle and plain README by @av in #2
• UI: from-scratch rewrite on a cursor.com/brand design system by @av in #12
• tracequest 0.2.0 by @av in #13

New Contributors

• @av made their first contribution in #1

Full Changelog: v0.1.0...v0.2.0

tracequest 0.1.0

First public release.

tracequest reads the session logs that coding agents already write to disk (Claude Code, Codex, Cursor, OpenCode, Grok, Droid) and turns them into a local UI: per-session timelines, tool calls and errors, token and cost breakdowns, and full-text search across all your history. It can also launch and steer agent runs through tmux, and pull sessions from other machines over SSH.

Try it

git clone https://github.com/viktor-com/tracequest.git
cd tracequest
npm install
npm run demo        # sample sessions at http://localhost:7777

npm start browses your own sessions. Needs Node.js 22.5+. Rust is optional (faster indexer).

Or install a prebuilt tarball from the assets below:
npm install -g ./tracequest-0.1.0-<platform>.tgz (macOS arm64, Linux x64).

Defaults worth knowing

• Localhost only. serve listens on 127.0.0.1. It has no login and can launch agents, so read SECURITY.md before using --bind. To share the hub, use a private network (Tailscale, WireGuard) or an authenticating proxy; see docs/hub-deployment.md.
• No telemetry, no unprompted network calls. The only outbound calls happen on request: share, import cursor-cloud, import ssh, limits, and serve --usage-limits.

What's in it

• Session browser and per-session view: timeline, chapters, tool usage, errors, retries, cost
• Search and filter expressions across every source (tool:Read AND error, host:gpu, …)
• insights: cross-machine view of where agents fail and wait
• Hub mode (serve --hub): read-only, secret-redacting view for a team
• Launch, steer and continue agent runs from the browser
• Optional Rust sidecar for indexing large histories

MIT licensed. Not on npm yet.