Background Story
Previously, I was using cmux to run multiple AI sessions across providers. I've Claude Max (20x) & Codex Pro (5x) subscriptions. The reason is that I never trust only 1 single ai models to do my job. It's like when u have a company, and it only has 1 employee, and that person can say any bullshit to you. In the AI model case, the AI itself can hallucinate.
So, in cmux i need to copy-paste manually from 1 window terminal (claude) to another window terminal (codex) just to make them work together. It's very painful, and also they cannot have the same state of memory and knowledge.
A Solution from me
I built an AI Agentic IDE from scratch using Rust and Tauri. Not a VSCode fork. And there are a lot of features that I, as a developer, needed.
Connect Local CLI
Connect through my Local CLI installed (e.g., Claude Code CLI, Codex CLI, OpenCode CLI). This ensures it will use the usage from my monthly subscription and not api usage.
Custom Agent Persona
This feature lets me assign a Name, Role, and specific System Prompt for any AI models across providers that fit with my project.
Multi Agent Orchestra
Finally, I can do an orchestrator on all AI models across providers in 1 chat window conversation without needing to copy-paste manually again. And the orchestrator model actually can do an orchestration. Like assigning them a task, reviewing other AI agents' work, coordinating with them, and many more.
Project Context
The project context that is available in this IDE is not only about codebase indexing. There are other contexts like Knowledge Base, Rules, Memory, and the newest one is Wiki Engine.
Kanban Board
Like in professional work, we usually break down any job into tasks as small as possible. This will ensure everyone (now every ai agents) can work in parallel at the same time. Each task will have its own git worktree.
Built-in Docker Container
Docker Desktop is heavy as f*ck, and OrbStack is only available on Mac OS. So, I built my own optimized Docker engine that can run on multiple operating systems (mac, windows, linux).
Canvas
Why should we subscribe to another tool just for designing? In this new AI Agentic era, it would be better if everything can be done in 1 place, and you are also able to use your own subscription to do a design. Once finished, bring it to the main Chat to ask the AI to build it.
In-App Browser
If you're building a website or web app a lot. The built-in app browser will help you a lot. You can annotate any specific DOM element to do a revision or changes, mention the network log if there is any error, checking if it's responsive or not. Without leaving the IDE, meaning you will not need to open a browser like Chrome that can consume a lot of resources
Remote SSH Manager
Connect & manage any remote SSH inside the IDE. You can connect to staging, qa, and any VPS instance easily without needing to remember all the credentials. And it's fully sotred end-to-end encrypted.
Wiki Engine
Having a large codebase project? Or u just got handover for legacy project whose documentation sucks? No worries, use the wiki engine to do full mapping on your project, and it will produce pages and the graph for you and your AI Agents.
Public Beta Test
Finally, after 3 months of development and a Private Beta Test. Now i bring it to the public so you can use it in the current early access stage for FREE !!! So, what are you waiting for? Try it now https://sokudo.dev/
Top comments (0)