You, billing your client
Git commits, Claude Code sessions, container runs — tracked automatically. Review at month-end, send invoice.
Your environment already knows what you worked on. Tardis transforms your daily engineering signals—commits, container runs, and active project directories—into precise billable hours in the background.
Whether the work is done by you or by an agent you deploy — Tardis tracks it, meters it, and drafts the invoice. Armed with a production-ready MCP server and standard REST webhooks, Tardis gives your terminal AI agents full CRUD control over your timesheets, while broadcasting real-time automation events straight to your engineering stack. Let your agents manage logs, rewrite entries, and draft invoices seamlessly. Stop fiddling with stopwatches.
You tell yourself you’ll log it later. By Friday you’re reconstructing Tuesday from memory — guessing, rounding, and leaving money on the table. Every timer-based tracker has this problem. Discipline shouldn’t be a required feature.
Tardis watches your git commits, Claude Code sessions, and container runs from the moment you start. At month-end: review what it captured, approve, send the invoice. No stopwatches, no reconstruction, nothing lost.
And because it ships a native MCP server, your AI agent can log hours directly too — useful when Claude Code is running a long refactor in the background while you’re on a call with a client.
The same logic applies when the worker isn’t human. A coding agent running overnight for a client, a deployed bot processing documents, an AI-powered service billing by the task — Tardis gives every actor, human or agent, the same metering and invoicing layer.
Git commits, Claude Code sessions, container runs — tracked automatically. Review at month-end, send invoice.
Human hours and agent sessions on the same timesheet. One client, one invoice, full breakdown.
Deploy an agent with a Tardis account. It tracks its own sessions and generates invoice drafts — no human in the loop except for approval.
Your users buy agent-hours, tokens, or task completions. Tardis is the metering layer. You connect your billing tool.
One intelligence engine reconstructs your day from whatever you give it — deterministic signals, plain language, or a direct call from your agent.
Tardis reads deterministic engineering activity — git branch, commit and diff density, AI-agent prompt volume, local container runs, macOS window & idle state — and reconstructs focused work sessions automatically.
no stopwatch, no guessingAnyone can brain-dump in plain language, typed or spoken. The engine unpacks it into clean, dated time slots in your timezone — no form fields, no categories to pick.
“45 minutes on accounting for client X this morning, an afternoon call with the lawyer for client Y, then ~2 hours on a deck for client Z.”Tardis ships a native MCP server and standard REST webhooks. Agents read and write slots directly — human or automated, the interface is the same. First-class client, not an afterthought.
built for agentsTime flows in from anywhere, through one engine, and back out to anything you own. Web chat, API, MCP, bots, and webhooks all drive — or are driven by — the same brain.
One engine
resolves · de-dupes · provenance
Every entry shows where it came from — ambient, chat, an agent, or a bot. Honest provenance, one source of truth, no black boxes.
Two new surfaces extend the same engine — message Tardis from the apps you already live in, and pipe its events into anything you run.
Track time from wherever you already chat — no context switch. Telegram & Slack both live, running the same brain as web chat.
Personal tracking: you link your own account, even inside a shared Slack workspace. WhatsApp comes later; voice notes are a phase-2 add.
Tardis is a platform you can build on, not a closed app. Register your endpoints and it calls them the moment something happens.
Draft-only holds: invoice.draft_created means a draft exists. There is no “sent” event — Tardis never sends one for you.
Both ambient slots bill the full wall-clock minute they ran — no fractional-share math.
You drive project A while a test suite, build, or background agent runs on project B. Tardis is built to track independent slots concurrently — each sustained by its own stream of signals.
Every slot lives or dies by its own activity — so two real workstreams never get collapsed into one fuzzy timer.
When a project goes quiet, its ambient slot closes itself and is flagged for your review — never dropped behind your back.
Phone calls and meetings live on a separate manual project you start and stop by hand. Deliberate work, deliberate control.
Say “25-minute call with Mr. Boninger” and Tardis resolves Boninger to the right client from what it has learned before. New name? It asks once, then remembers the answer forever.
At month-end, your tracked time becomes a ready-to-review draft in lexoffice or SevDesk — or a clean CSV. Grouped by client and project, with your rate and your wording.
Tardis is built for a privacy-sensitive, EU-facing audience. Every signal source is opt-in, off by default, and revocable any time — and you can read the code that does the capturing.
No listener runs until you turn it on. Each source — git, agents, containers, window state — is a switch you flip, and can flip back.
Turn a source off and capture stops immediately. Nothing leaves your hands on its own, and nothing keeps watching once you say stop.
The macOS menu-bar app is open source, so you can verify exactly what’s captured — line by line, not a paragraph in a privacy policy.
read the source — coming soonExport everything — time entries, clients, projects, invoices — as JSON or CSV, or pull it all via the MCP server or REST API. No tier gate, no waiting period, no sales call. If you leave, you leave with everything.
Manual time tracking is unlimited on every plan, including Free. What’s metered is the agent — the part that actually costs us compute.
A real tracker, not a crippled demo. Unlimited manual entries, forever.
no credit card · no trial clock
For the engineer whose agent should draft the month’s invoice.
€5.60 / mo billed yearly · save 20 %
Unlimited seats. Pay for work done, not heads — onboard everyone without thinking twice.
€20 / mo billed yearly · unlimited seats
Buy extra AI requests in one-off packs. Purchased requests roll over and never expire — your monthly allowance is spent first, and a pack only draws down once you’ve used it up.
Need SSO, a custom request volume, or a contract?
Let your environment log the work you’re already doing.