Summary of "OpenCode Desktop: Great UI/UX + Run Multiple AI Agents (Free)"

What Open Code is

Open Code is an open-source, AI-powered coding tool that combines a CLI with a new desktop companion.
It connects to multiple AI model providers, letting you choose closed- or open-source models.
Positioned as an alternative to tools like Cloud Code, Codeex, and Gemini CLI.

Built as a fast, lightweight desktop companion using Rust and Tauri (transcript references a similar framework name).
Modern UI/UX: dark theme, clean terminal UI, workspace/session metaphors, and terminal tabs.
Editing and review features:
- LSP support and inline file viewer.
- Clean diff views with unified/split options and inline comments.
- One-click to open files in external IDEs (e.g., VS Code, Cursor).
Multi-session/workspace support: run multiple AI agents in parallel on the same or different projects.
Integrates with the Open Code CLI — you can run the CLI from the app.
Per-session model selection and per-model “thinking effort” presets (default / high / max) for some providers.
Ships pre-configured examples of free models (video mentions GLM5, Kimik K 2.5, and a higher-tier model used in the demo).

Getting started
- Download the desktop app from the Open Code website, install, open, configure appearance, and connect providers/models.
Create a new project (data-analysis web app)
- Use Plan mode: the agent asks clarifying questions (HTML templating, chart library, client-side CSV processing, level of statistical insights).
- Agent produces a plan; switch to Build mode to let it write files.
- Result: a working web app with CSV upload, client-side parsing, summaries (rows/columns, numerical vs categorical), column insights, trends, correlation matrix, and interactive charts (bar/line/pie/histogram) with X/Y selection and chart-type switching.
- Use inline comments to direct the agent to specific code areas; view generated files and run locally.
Work on an existing project (Next.js personal finance app)
- Open the repo, run the dev server in the app’s terminal.
- Task the agent to add a “goals” page: create DB tables, reuse styles/components, add create/edit goals and contributions.
- Agent edits files; changes are highlighted and viewable via diff tabs. The demo shows a functional goals page with goal creation and contribution balance updates.
Parallel sessions
- Launch multiple sessions (e.g., building the goals page while adding a dark theme to another project) and let agents work concurrently.

App is in beta — expect UI bugs (example: a to-do list rendered twice).
Agent behavior issues observed:
- Agent remained in Plan mode and initially could not write files until the user manually switched to Build mode.
No built-in Git GUI integration — commits/push/pull require using the terminal (users have requested this).
Lacks some advanced automation features present in Codeex desktop (per the reviewer).

Comparable to Codeex desktop in core idea (AI-assisted project editing, diffs, inline comments) but currently missing some automation and Git integration that Codeex may offer.
Overall impression from the video: fast, beautiful, and responsive.
Recommended as a companion for existing Open Code CLI users — particularly valuable for its multi-session workflow.

Good fit if you already use Open Code CLI and want a GUI for multi-session AI-assisted development, diffs, and inline guidance.
Use Plan mode for clarifying prompts before building; switch to Build mode to allow file writes.
Expect beta quirks; watch for upcoming updates and requested features (notably Git integration and improved build-mode automation).