# FlowSight: plan a session, protect your focus, and review your work locally > Source: > Published: 2026-10-12 10:31:43+00:00 I'm Manuel Castillejo, the creator of [FlowSight](https://solo.flowsight.site/?utm_source=devto&utm_medium=community&utm_campaign=solo_first_session). It's a desktop application for planning a work session, keeping track of it, protecting a focus block, and reviewing what happened with local AI. A common example is a task that takes you through an editor, documentation, a browser, and tests. That activity can help you reconstruct the session. You still need to compare it with the actual result: a finished change, a document, or a problem that remains unresolved. FlowSight brings those steps into one workflow. *The application interface shown above uses illustrative demonstration data.* Choose the task you want to work on and the time available. The local planner can suggest work blocks using your tasks, preferences, and recorded task time. You review the plan before applying it. You can use the normal timer or Pomodoro. The practical starting point is one session that answers a small question: did the block help you make progress on the task you chose? Total Focus connects the session to Browser Controls. You can block distracting websites or paths while keeping exceptions that the task needs. For example, a tutorial might be useful during a coding task even when other browsing is distracting. The controls let you make that distinction. Browser protection requires the extension to be installed and paired. Insights includes reports for today, the last seven days, and the last thirty days. You can review the report and export a PDF. The report identifies whether it used local AI or rules and shows limits such as how much recorded activity was available. If local AI is unavailable, the report uses identified rules. A report without activity shows that missing data rather than proposing a change from nothing. Recorded activity does not establish that a task was completed. Before using a summary in a work update, compare it with the files, decisions, or results it describes. A recommendation can become a decision to try, with a review date. You can also keep the current plan. Those review decisions are stored on the device and follow the local retention and deletion settings. The useful outcome is a small change that you can evaluate after another session. The local edition includes a model for analysing screen context on the device. Local tracking and reports work without a cloud account. Optional cloud services and external connections are separate choices. For example, a cloud AI client that requests report data through MCP can send that data to its provider. The product's local processing does not make every optional integration local. FlowSight also provides controls for when tracking runs, excluded applications, saving window titles, and retaining local history. The examples here describe the Windows edition. [FlowSight 6.1.1 has published Windows EXE and MSI installers](https://github.com/Mancasvel/FlowSight.AI/releases/tag/v6.1.1). macOS and Linux releases follow their own update schedules, so check the release for your platform. The [source is available under AGPL-3.0](https://github.com/Mancasvel/FlowSight.AI). You can use, build, modify, and redistribute it under that licence. The Solo website currently describes a planned EUR 25 one-time official distribution offer, with checkout not open yet. Previously published installers remain available during the transition. Optional cloud plans are separate. To try the work-review flow: install the relevant release, start tracking for a work session, and open **Insights ? Work report**. Check the report against what you actually produced. If you try it, I'd be interested in the point where the session-to-report workflow becomes unclear, and which information makes the report useful for your work.