Files, knowledge, contacts, monitors, and a publish button. Backed by a real folder on your disk — not a database you can’t open with another tool.
Every workspace is a folder. Every section is a subfolder. Every item is a file. You can open it in Finder, edit with another app, sync via your preferred tool, and Egg will pick up the changes.
No proprietary database for project content. Open the folder anywhere — Finder, Explorer, terminal, another editor — and your data is there.
Toggle between the workspace UI and the raw file manager. Same data, different lens.
Drag files into sections. Reorder by drag. Sections persist as folders on disk.
Markdown, PDFs, images, videos, captured web pages, monitor reports, notes — whatever you drop in.
Right-click a tab or page → Add to Project. Captures URL, title, body, screenshot, capture timestamp.
Right-click any link to add it to a project without navigating.
Hit "Compile knowledge." The agent reads every file in the project and writes a synthesis to knowledge.md. Re-runs when you ask. Marks itself stale when monitors bring new data.
Run the knowledge skill from the project page. The agent reads every file, builds a synthesis, writes it back to the project as Markdown.
When a monitor or new file changes the project, the knowledge view shows a "stale" badge so you know to re-run.
Keyword search across every text file in a project. Hits get inline previews with surrounding context.
The output is a Markdown file you can edit, version, copy, or share. Not a black-box vector store.
Web monitors check URLs on a schedule, detect changes, and write reports straight into the project folder.
Visit a URL, point at what should be tracked, and the LLM derives the extraction logic. No JSON config.
Set the interval — minutes, hourly, daily. Monitors run in the Gateway daemon, even when the app is closed.
Each detected change writes a Markdown report to project/monitors/ with the diff, timestamp, and source URL.
Telegram, email, or system push when conditions match. Per-monitor channel routing.
"Notify when price drops below X." "Notify when stock changes." "Notify on any change." LLM-derived from natural language.
Egg ships a CRM-grade contact system without acting like one. Every person you research becomes a profile — notes, memories, social links, work history, education, recent activity.
Open any person’s page (LinkedIn, personal site, Twitter) and click Extract profile. Egg parses and saves a contact.
Free-form notes and structured memories per contact. The agent surfaces them automatically when you mention the person.
LinkedIn, Twitter, GitHub, personal sites — tracked per contact and used as launch points for re-extraction.
Career and education history extracted automatically. Editable.
Flexible tagging for people and organizations. Filter contacts by tag.
You don’t "use" a CRM. Egg builds the contact graph as a side effect of normal browsing.
Turn any project into a public canvas with a shareable URL. Edit in place. Copy the link. Unpublish when you’re done.
Lists every published canvas, with create-new, edit, copy-link, and unpublish actions.
Start a canvas from an existing project or build one from a blank page.
Inline rich-text editor with embedded artifacts, screenshots, and links. The same editor that backs project knowledge.
One click to copy. Anyone with the link can read. No account required to view.
Pull the canvas off the public site without deleting it locally.
Each workspace gets its own agent surface — specialized for discovery, research, organization, and monitoring inside that project.
A dedicated agent rooted in one project. It knows the files, the contacts, the monitors, the knowledge state.
Skill loadouts per workspace agent. Toggle which skills apply to this project — research-only here, organize-only there.
The /agent route is a spatial canvas. Pin nodes (research questions, contacts, articles) and the agent walks the graph.
The agent itself can add files, contacts, and monitor outputs to the right project — routed by intent.
P2P encrypted chat ships in the backend today. The UI is parked while we get the basics right.
Cloud-relay DMs over the same E2E-encrypted transport that powers sync. Phase 1 active in the daemon.
Backend onlyThe agent has full Gmail access via Google Workspace tools. Standalone email UI is parked — for now, you read mail in Gmail and ask the agent to act.
Backend onlyFree to download. Bring your own API key.
Download for WindowsAlso available for macOS