one-time
For one person on up to Macs.
- All 26 agents
- Every feature
- Offline license, no account
Every coding agent’s sessions,
on one deck.
HyperTrail finds the AI coding agents on your Mac and turns their history into live dashboards: every transcript, token, dollar and usage limit. There’s nothing to set up and nothing to sign in to.
days, every feature·No account, no card· or later
Agents26 supported
Twenty-six agents. Nothing to configure.
HyperTrail reads each agent’s own history folder, read-only. Only the agents on your Mac get a tab, and each tab wears its vendor’s colors.
Claude CodeAnthropic
CodexOpenAI
Gemini CLIGoogle
Qwen CodeAlibaba Qwen
GitHub CopilotCLI, VS Code & JetBrains
CursorEditor & Agent CLI
WindsurfCognition
AntigravityGoogle
OpenCodeAnomaly
CrushCharm
AiderAider AI
GooseBlock
AmpAmp
DroidFactory
Kimi CodeMoonshot AI
ClineCline
Roo CodeRoo Code
Kilo CodeKilo
ContinueContinue
PiTerminal agent
KiroAWS
Zed AgentZed Industries
WarpWarp
JunieJetBrains
AuggieAugment Code
OpenHandsAll Hands AI
Every surface counts: Copilot in the CLI, VS Code and JetBrains; Codex in the CLI, IDE extension and desktop app; Cursor’s editor and Agent CLI; Kiro and Antigravity in the IDE and the CLI.
01Dashboards
Every session, as a card you can act on.
Title, project and branch, when it started and how long it ran, the time the agent actually spent working, which models, how many tokens, and what it would have cost at API prices.
Card style
Lite mode
The leaf in the toolbar trims every card to its title, project, one line of numbers and Resume.
Filters that matter
Project, source, sort order, named sessions only, and whether to include SDK and subagent runs.
Live at a glance
A pulsing dot while an agent works or waits on a tool. Blue when the session is open and waiting for you.
Grid or list
A compact list for hundreds of sessions, and bulk select when it’s time to tidy up.
02Conversations
Read the whole conversation, not just the prompts.
Replies, tool calls paired with their results, thinking, and subagents rolled into their parent session, rendered as Markdown with syntax highlighting.
- Declutter
- Hides step-by-step tool noise. One click brings it back.
- Find
- ⌘F with a live match count; ⌘G steps through.
- Follow live
- Stays on the newest message while the agent writes.
- Prompts
- Every prompt you sent, filterable, copyable in one go.
- Artifacts
- Each file the session wrote or edited, with red and green diffs.
- Insights
- Token mix, cost per model, tool use and how the context grew.
- Export
- Any conversation as clean Markdown.
03Usage limits
Know you’ll hit the limit before you hit it.
Live usage straight from each vendor: Claude’s 5‑hour and weekly windows, Codex, Copilot premium requests and Gemini’s daily quota. Each comes with a reset countdown and a pace check.
Pace, not just percent
The tick on each bar is where a steady pace would put you right now. Ahead of it, HyperTrail tells you the window runs out before it resets.
The 5‑hour block, and a heads-up
For Claude Code it also works out the current 5‑hour block from your own history, ccusage-style: tokens, cost and time left. You get a notification when a session finishes or a limit passes 80%.
04Usage analytics
What every agent cost, by model and by project.
Tokens and API-equivalent cost across all your agents, from the last 24 hours to the last year. It’s all computed on your Mac from session history and priced from a bundled table of 1,254 models.
05Menu bar
Your tightest limit, always in the menu bar.
The menu bar shows whichever limit is closest to running out, or the agents working, or today’s cost. Click it for every limit, the current 5‑hour block, the last seven days and what’s running now.
- Shows
- Tightest limit, agents working, today’s cost, or just the icon.
- Live
- Running sessions from every agent, one click from their transcript.
- Quiet
- Hide the Dock icon and keep HyperTrail in the menu bar alone.
06Search
Every session in every agent, one keystroke away.
⌘K searches titles, prompts, paths and branches across all your agents as you type. This one works, so try it.
Search every session⌘K
Find in a dashboard or conversation⌘F
Next / previous match⌘G ⇧⌘G
Jump to an agent⌘1–9
Overview⌘0
Usage⇧⌘U
Back to the dashboard⌘[
Refresh⌘R
Settings⌘,
07Actions
Pick up exactly where the agent stopped.
Every card carries the agent’s own resume command. Everything else is a click away.
Privacy & security
Local by design. Read-only by default.
HyperTrail works from the files your agents already keep. There’s no account, no server of ours and no telemetry, and it never writes to another app’s data.
On your Mac
- ~/.claude/projects
- ~/.codex/sessions
- ~/.gemini/tmp
- …/Code/User/workspaceStorage
- …/Cursor/User/globalStorage
- ~/.local/share/opencode
- and each other agent’s own folder
read-only
HyperTrailKeeps its own index
in Application Support
in Application Support
Leaves your Mac: only these
api.anthropic.comClaude limitsopt-inchatgpt.comCodex limitsapi.github.comCopilot limitscloudcode-pa.googleapis.comGemini limitssessiondeck.appUpdate checkraw.githubusercontent.comPrice tableon request
Each one can be switched off in Settings. Limits calls carry only the login that agent already saved, to that agent’s own vendor. No session content is ever sent.
Read-only
Agent history is opened read-only, SQLite stores included, and never modified. HyperTrail doesn’t write to another app’s database.
Changes only when you ask
A rename is written in the agent’s own format, keeping file timestamps. A delete goes to the Trash, and is refused while an agent is working.
No credentials to type
Live limits reuse the login each CLI already saved. Claude Code’s Keychain item is read only after you opt in, and macOS asks you once.
Tokens stay in memory
Access tokens go only to that vendor’s own usage endpoint. They’re never written to disk, logged or shown.
No telemetry
No analytics, no tracking, no account. The requests above are the complete list.
Offline licenses
License keys are Ed25519-signed and verified on your Mac with CryptoKit. There’s no license server and no activation call.
Performance
Native, and fast where it counts.
Swift and SwiftUI with incremental parsing. Screens you’ve visited stay alive, so coming back costs a frame, not a rebuild.
1.0s
to index 888 MB of Claude Code history from cold: 416 sessions, subagents included.
0.02s
for a warm rescan of everything.
<40ms
to switch back to a tab you’ve already visited.
20ms
95th-percentile frame time, flinging through a 900-card grid.
≈0%
CPU when idle, and about 5% while agents stream to disk.
~1s
from an agent writing to disk to its card starting to pulse.
Off the main thread
Scans and every aggregate run in the background, and each result lands in one update.
Screens stay alive
Visited tabs are kept and frozen; the likely next ones are prewarmed.
Cheap cards
Each row of chips and buttons is drawn as one view, which makes cards three times cheaper to build.
Search without stutter
A tight scan over pre-lowered text: no frame over 40 ms while you type across 900+ sessions.
Measured on an M-series Mac with 26 providers and about 1,000 sessions, with an agent streaming the whole time.
Pricing
Try everything. Then pay once.
Start with days of every feature, with no card and no account. Keep it with a one-time license, or put your whole team on it.
Best value
one-time
For one person on up to Macs, for good.
- Everything in
- All future agents
- Lifetime updates
- Priority support
per seat / year
For teams of or more, on one invoice.
- Everything in
- A license key per teammate
- Reassignable seats
Seats
per year
Buy for your team
Custom pricing for larger rollouts.
- Volume licensing
- MDM deployment (pkg + managed license)
- Offline license files
- Security review pack (data flow, no telemetry)
- Priority SLA
- PO and invoice billing
Upgrade anytimeMoving up a plan? Only pay the difference.
-day refundNot for you? Get your money back, no questions.
Verified offlineLicenses are checked on your Mac. No account, no server round-trip.
Compare plans
| Feature | Free trial days | once | once | / seat / year | Custom |
|---|---|---|---|---|---|
| The app | |||||
| All 26 current agents | |||||
| Every feature | |||||
| Agents added later | During trial | Not included | |||
| Lifetime updates | Not included | Not included | While subscribed | ||
| Licensing | |||||
| Macs | 1 | Per seat | Volume | ||
| Verified offline, no account | |||||
| A key per teammate, reassignable | Not included | Not included | Not included | ||
| One invoice | Not included | Not included | Not included | PO & invoice | |
| MDM deployment and offline license files | Not included | Not included | Not included | Not included | |
| Security review pack | Not included | Not included | Not included | Not included | |
| Support | |||||
| Email support | |||||
| Priority support | Not included | Not included | SLA | ||
Which agents does HyperTrail support?
Claude Code, Codex, Gemini CLI, Qwen Code, GitHub Copilot (CLI, VS Code and JetBrains), Cursor, Windsurf, Antigravity, OpenCode, Crush, Aider, Goose, Amp, Droid, Kimi Code, Cline, Roo Code, Kilo Code, Continue, Pi, Kiro, Zed Agent, Warp, Junie, Auggie and OpenHands.
How much each tab can show depends on what the agent stores. Most keep full transcripts; a few, such as Windsurf’s encrypted Cascade history, only expose part of it, and HyperTrail shows what’s readable.
Do I need to set anything up?
No. HyperTrail finds each agent in its standard data folder and honors variables like $CODEX_HOME. If you keep an agent’s data somewhere unusual, point HyperTrail at it in Settings.
Does it need my API keys or passwords?
No. You never type a credential. For live limits, HyperTrail reuses the login each CLI already saved on your Mac. Claude Code keeps its login in the Keychain, so HyperTrail asks you first, and macOS asks once.
What leaves my Mac?
Only the requests listed under Privacy: each vendor’s own usage endpoint for live limits, the update check, and the price table when you ask for it. Session content never leaves your Mac, and every one of those requests can be switched off.
Will it change my agents’ files?
It reads them read-only. It writes only when you ask: a rename is saved in the agent’s own format, and a delete moves the session to the Trash. Deleting is refused while an agent is still working in that session.
How is cost calculated?
From token counts at API list prices, using a bundled table of 1,254 models from LiteLLM that you can refresh in Settings. Where an agent records its own cost, HyperTrail uses that and marks it with a check. On a subscription you aren’t billed per token, so treat it as a measure of value.
What happens when the trial ends?
Your first agent’s dashboard stays free. A license adds every other agent, usage analytics, notifications, resume in terminal and export. Nothing is deleted either way.
Can I use one license on several Macs?
covers Macs and covers . gives each teammate their own key, and you can reassign seats as people come and go.
How do licenses work without an account?
Your key is signed with Ed25519 and verified on your Mac with CryptoKit. There’s no account to create, no license server, and nothing to activate online.
Can we roll it out across a company?
Yes. Enterprise includes an installer package and a managed license for MDM, offline license files, a security review pack covering the data flow and the no-telemetry policy, a priority SLA, and PO or invoice billing.
What does it run on?
or later.
See everything your agents have been doing.
Free for days. No account, no card.