Meridian

Meridian is an open-source desktop app built by Meridiona that automatically reconstructs your workday from screen activity, keeping a private, searchable journal of everything you worked on — without any manual input. It runs entirely on-device (macOS Apple Silicon or Windows 10/11), is MIT-licensed, and is free to use. The project reached v1.90.0 as of August 2026 and has accumulated over 250 GitHub stars.

What It Is

Meridian is a local-first work journal and time tracker that watches what you do on screen, organizes it into a timeline, and produces daily summaries and ready-to-post worklog updates for project management tools. It is explicitly positioned as "not a monitoring tool" — all capture, analysis, and storage happen on your device, and nothing is uploaded without your explicit approval. The core product is written in Rust (with a Tauri tray interface) and is fully auditable under the MIT License.

How It Works

Meridian runs quietly in the background, capturing the apps and windows you work in throughout the day. At the end of the day it:

• Reconstructs a timeline of tasks, switches, and time spent per activity
• Generates a daily summary explaining what got done, what ran over estimate, and why
• Drafts worklog updates matched to the right project tickets
• Writes a standup note ready to paste or post with one click

The AI analysis step sends only what a summary needs to whichever AI provider you connect (e.g., Claude, ChatGPT, Gemini); the raw screen data never leaves your machine.

PM Tool Integrations

Meridian connects to project management tools in a read-only mode until you explicitly approve a log post. Supported integrations include:

• Jira (available now)
• GitHub Projects (available now)
• Linear (coming soon)
• Azure DevOps (coming soon)
• Trello (coming soon)

The tool is also fully useful without any PM integration — it functions as a private, searchable work record on its own.

Privacy and Architecture

Privacy is described by the project as "the foundation, not a feature." Key architectural decisions include:

• All capture and analysis run on-device
• Activity is stored in an encrypted local database
• Diagnostics are opt-out and stripped of identifying information before leaving the device
• Any app can be excluded; capture can be paused from the menu bar at any time
• The MIT license means every line of code is publicly auditable

System requirements: macOS (Apple Silicon only; Intel not supported) or Windows 10/11; 0.5–3 GB RAM; approximately 20 GB storage per month (older captures pruned automatically after 30 days).

Update: v1.90.0

The latest release is v1.90.0, published on August 25, 2026. The repository was last pushed to on August 27, 2026, indicating active development. The project launched on GitHub in May 2026 and has grown to 253 stars and 21 forks within its first few months, with 18 open issues tracked publicly. The homepage features a Product Hunt top-post badge, signaling early community traction.