Skip to content
session intelligence · in ronda now

See where yourtime goes.

Ronda keeps every session from your coding agents, and learns from them on your machine: where the hours go, where agents get stuck, the errors that keep coming back, and the stack you actually build with.

download for mactry the report

Open Intelligence in the top bar, or press ⌘3 · read the docs

the real view · sample data · change the range and project
loading the report…
questions

Answers your session history already holds.

Where did the time go?

Active time split into features, bug fixes, refactoring, tests, setup, and exploring, plus the time spent recovering from failing commands. Breaks longer than 15 minutes don't count.

Where do agents get stuck?

Edit loops, context that filled up and was compacted, tests still failing at the end, denied tool calls, and calls to APIs that don't exist.

Which errors keep coming back?

Error lines are normalized so the same bug matches across sessions, agents, and projects, and flagged when it returns after a session that committed a fix.

What is my real stack?

Languages, frameworks, and tools read from the files your agents edit and the commands they run, with anything new in the period marked.

How do sessions end?

Committed, edited but not committed, ended failing, or no file changes, with the time each outcome took.

Where's the proof?

Failures and recurring errors link to the sessions behind them. Click one to open the transcript at the message where it happened.

how it works

Same rules as the rest of Ronda.

Filter any report to the last 7 days, 30 days, or all time, and to a single project.

  1. 01

    Worked out while indexing

    Ronda derives facts from each transcript as it indexes it: tool calls, errors, edited files, commands, and outcomes. Reports read those facts, so they stay fast on large libraries.

  2. 02

    Deterministic and local

    Rules, not a language model. Nothing leaves your machine, and the same sessions always give the same report. Agents that don't record tool calls count toward time and sessions only, and the view says which ones.

  3. 03

    Ask from any agent

    ronda-mcp exposes ronda_insights and ronda_find_error, so an agent can check whether an error has been seen before it starts fixing it again. ronda-cli insights and ronda-cli errors do the same from a shell.

What should it notice next?

Tell us which patterns would save you the most time, or where a report got your sessions wrong.

share an ideaget ronda