Native time tracking for your AI agents

Why track your AI agents' time?

Agencies delegate more and more work to AI agents: writing code, drafting content, running research. That work is real, but it stays invisible — nobody starts a timer for an agent, so hours pile up unbilled.

Rytmely turns that hidden effort into billable time. Every task your agents run is logged automatically, giving you an accurate view of what was produced and what to invoice your clients.

How it works: the MCP protocol

MCP (Model Context Protocol) is an open standard that lets AI agents talk to external tools. Rytmely exposes a native MCP server, so your agents log their time directly, with no third-party gateway like Zapier in between.

• start_timer — the agent starts a timer when it begins a task
• stop_timer — the agent stops the timer once the task is done
• list_active_timers — see every timer currently running
• list_projects — the agent picks the right client and project

Humans and AI agents tracked in one place

A human keeps a single active timer at a time; an AI agent can run several timers in parallel, one per concurrent task. Rytmely handles both models natively without any manual workaround.

Everything lands in the same Client → Project → Label structure, so your reports blend human and agent time seamlessly and stay ready to invoice.

MCP-native, European and GDPR-ready

Rytmely connects directly to Claude Desktop, Claude Code and Cursor through native MCP — no fragile third-party bridge. Your setup stays simple and reliable.

Your data is hosted in Europe and GDPR-compliant, with a clear French-and-English interface. MCP is included from the free plan (200 calls per month, one API key on Solo).