Time tracking for AI-augmented developer workflows
Problem
Freelance developers who bill by time find that half their work is now prompting AI agents, reviewing their output, and waiting for them to finish. Classic per-app time trackers can't capture this honestly, so people resort to eyeballing hours.
Opportunity
A time tracker purpose-built for AI-assisted work: detects prompt/review/wait cycles across IDE, terminal and agent tools, and produces honest, billable breakdowns of human vs. AI-assisted time.
Market analysis
The pain is authentic and loudly expressed by freelancers, and several small tools already attack it — but they are free or open source, which signals demand while capping monetization. Winning requires winning the billing-trust angle, not just the tracking.
Market · Freelance and contract developers billing hourly while working with Claude Code, Codex, Cursor and similar agents.
Pricing · Direct comparables are free/open source (agent-hours, claude-code-time-tracker) or free-tier local apps (trackmy.codes), leaving little pricing power; classic trackers anchor around low per-user monthly fees.
Pros
- + Session logs of coding agents are machine-readable, so the human/AI/wait split is genuinely computable.
- + A hands-on vs supervised vs autonomous breakdown is exactly what hourly clients will start demanding.
- + Cheap to build as a local-first tool reading existing agent logs.
Cons
- − Multiple free/open-source tools already cover the core tracking use case.
- − Freelancers are a small, price-sensitive consumer-ish market with high churn.
- − The billing-honesty framing invites client distrust if the tool, not the freelancer, classifies time.
Existing / similar tools
Source
Hacker News (Ask HN)
Search surfaced at least four tools already doing variants of this, including one (agent-hours) with the exact hands-on/supervised/AI-autonomous split the idea proposes — so the question is not feasibility but why none has broken out. The likely answer is that tracking alone is a nice-to-have; the paying wedge is invoice defensibility: a breakdown a client can audit and trust, maybe with agent-session receipts attached. That repositions the product from “another time tracker” to “billing evidence for AI-assisted work”, which agencies and consultancies could justify as overhead. As a consumer utility it dies to free CLI tools; as a billing-integrity layer it might sustain a small subscription.