DevClocked

DevClocked automatically measures the real work behind what you ship: human hours, agent runs, and token spend, captured at the source and tied back to the commits, PRs, and issues that actually landed. You ship with AI agents now, but nothing measures that honestly. Git-based analytics read the artifact after it lands and guess at the effort. Token dashboards read the bill and stop there. DevClocked sits inside the work session itself, across the IDE, terminal, agent runtime, and browser, and records what actually happened: - How long the human worked - Which agents ran - What they cost - What shipped as a result No timers. No manual logging. Set it up once and it catches everything, including the work that never became a commit. Key features - Zero-touch capture across every surface - Mac app, VS Code/Cursor extensions, Claude Code/Codex/CLI plugins, and - GitHub sync feed one normalized activity ledger. Nothing about your workflow changes. Leverage Score The headline metric: the ratio of shipped output to the human hours it took, measured across your work blocks and every agent you ran. It answers whether AI is a force multiplier for you or just expensive autocomplete. Time Slice Minute-level breakdown of real coding vs. config, debugging, review, and plumbing. The operational signal Git can never show. Agent monitor & tokenomics Every agent run is tracked per session with model, tokens, and dollar cost, broken down by project. AI spend ties directly to delivered output instead of sitting in an unexplained vendor bill. Issue-level outcomes Work blocks, agent runs, token spend, commits, and PR status roll up to Linear and Jira issues, so every ticket shows the hours, cost, and ship-state behind it. Multi-repo projects, teams & agencies Projects span repos. Teams get shared dashboards and leverage benchmarking with privacy-safe aggregation. Agencies get automatic timesheets and client-ready invoices covering both human and agent hours. MCP server Query your live signal, including sessions, focus, agent runs, spend, and shipped output, straight from Claude or any MCP client. What makes it different Everyone else reads the artifact. DevClocked measures the work. Real human hours are measured at the source rather than inferred from commit timestamps Concurrent agents across multiple repos appear in one view Per-run agent tokens and cost are captured rather than vendor-aggregated Work that never produced a commit still counts It's bottom-up from the developer's actual session, not top-down from Git history. That's the difference between: "45 commits this week" and: "4.2× leverage on 3 hours of real coding." Outcomes users can expect Developers prove their multiplier, not just their commit count A to-the-minute picture of the real day and the leverage behind every commit. Teams manage delivery on real work instead of keystroke proxies Benchmark leverage and attribute output across every engineer and agent. Agencies bill for everything they ship Defensible, evidence-backed invoices covering billable human and agent hours, generated without anyone filling in a timesheet. Everyone finds out whether their AI bill is compounding output See whether agent spend is creating leverage or getting burned on context resets and low-efficiency runs. Launch traction so far 10,000+ hours of source-captured work analysed 368B tokens tracked and mapped to projects 500+ multi-repo projects Plans Pro For solo developers. Includes Time Slice, Leverage Score, and token economics. Ultra Advanced mapping, exports, and unlimited history. Business Organisation dashboards, seats, and team benchmarking.