How do I track how long sessions take in Cursor (beta)?

Rius traces each Cursor chat (beta) as one trace, with timing and status on every span, so you can see how long a chat and each step in it took. Install the Rius plugin in Cursor, sign in, and turn tracing on for a folder. Then open the chat in the trace list in the Rius console. Cursor traces carry no token counts or cost, but they do carry the full span tree and timing.

How it works

The plugin turns each Cursor agent chat into a trace. It is built for the agent in the Cursor editor and cursor-agent in the terminal, though the beta is tested mostly on headless cursor-agent -p runs. Each trace has this shape:

• An agent span for the Cursor session at the root.
• A turn under it for each prompt and everything it caused.
• The agent’s answer under the turn, with its model.
• Each tool call, with a shell command that exits non-zero marked as an error.
• Each subagent, under the Task call that started it.

Every span records timing and status, so you can see where in a chat the time went. Spans arrive while you work: when the chat starts, after each answer, and every 20 tool calls. Span times are the moments Cursor reported each step, not the moment a model call started. Read them as the pace of the chat, not as exact model latency.

A chat’s trace closes when Cursor ends the session. If Cursor never says so, your next chat closes it once it has been idle for 12 hours. A chat counts in the figures on the Agents page once its trace is closed.

Timing is captured in both modes, structure only and with content, so you do not need to send prompts or replies to measure duration.

Set it up

1. Install the plugin. In Cursor’s plugin settings, choose Import from Repo and enter glassflow/rius-coding-agents. Installing sends nothing, and tracing stays off until you enable a folder. You need Python 3.9 or later on your PATH, bash, and macOS or Linux. Plugin version 0.6.0 or later is required for Cursor.
2. Sign in and turn tracing on. In the agent chat, run the commands below. /rius-login opens the browser, where you pick a workspace and click Connect. Run /rius-enable-here in the repo you want traced, which sends structure only: timing, status, models, tool names and error types, with no prompts or replies.
3. Check the state. /rius-status tells you whether the folder is traced and why.
4. Work as usual, then open the chat in the trace list at console.rius-glassflow.com.

/rius-login
/rius-enable-here
/rius-status

If cursor-agent sends nothing, check its version. Releases before 2026-08-27 ignore hooks that come from a plugin, and the Cursor guide shows how to add the hooks yourself.

To stop tracing a folder, run /rius-disable-here. A disabled folder beats any enabled parent. /rius-off turns tracing off for the current chat only.

What you get

• One trace per chat. Each Cursor chat appears in the trace list, and you open it to see every turn, answer, tool call and subagent in order, with timing on each span.
• Where the time went. Compare a slow chat with a quick one and find the tool call or subagent that stretched it. Failed tool calls are marked as errors.
• An agent named cursor. On the Agents page, chats show up as an agent named cursor, counted once each trace is closed.
• You by name. From plugin 0.6.2, every span carries your sign-in email as user.id, so your chats appear on the Users page. The Sent by column in the trace list shows whose key sent each chat.
• Questions in plain language. The plugin registers the Rius MCP server as rius. Click Connect next to it in Cursor’s MCP settings, or run cursor-agent mcp login rius, then ask about your recent Cursor chats.

What Cursor traces do not include: token counts and cost, because Cursor does not give plugins usage data. The plugin sends no heartbeat, so Cursor chats do not appear under Instances. Cursor’s cloud and background agents and Tab completions are not traced. Cursor support is in beta: it is tested mostly against headless (-p) cursor-agent runs, and tool calls in an interactive cursor-agent session, the agent in the editor and Import from Repo are not tested yet.

When something else fits

• If you need tokens and cost next to duration for each session, the same plugin reports them for Claude Code. See Track Claude Code usage and cost across a team.
• To compare how the coding agents differ in what they record, see Coding agents.

Related