RedSyl callme-sy

Snapshot

username: callme-sy
focus:
  - agentic orchestration
  - developer tooling
  - business-minded experiments
stack:
  - deepseek-harness
  - magpie
  - hermes-agent
  - opencode
  - command-code
currently:
  building: taste for useful systems and sharper workflows
  open_to: conversations, collaborations, and good product ideas

I tinker with agent workflows, tooling, and product-shaped ideas with a business bias. I care about systems that feel sharp, useful, and intentional, and I like the moment when an experiment starts looking like a real tool.

Featured

The five I actually keep open. Everything below is downstream of these.

DeepSeek Harness Everything is a plugin. The harness I build inside every day, and the reason my workflow looks the way it does.	Magpie Every agent's model in one place. Codex on DeepSeek, Claude Code on Kimi, switched from the menu bar instead of by hand-editing config.	Hermes Agent The agent that grows with you. Memory, self-improvement instincts, and a chat surface I can actually work with.
OpenCode The open source coding agent. Provider-agnostic by default, so I can move between models without rebuilding my setup.	Command Code A coding agent built for open models. What I reach for when the model is DeepSeek, GLM, or Kimi rather than a frontier API.

About

I am most interested in the layer where orchestration, product thinking, and execution meet. That usually means agents, workflows, human-in-the-loop interfaces, and the details that make tools actually usable.

This profile is less a portfolio of finished products and more a map of what I keep returning to while figuring out what to build next.

Signal

The Rest of the Stack

Everything that supports the five above.

Agent Harnesses

Tool	Signal	Why I keep it around
openclaw		Your own personal AI assistant that runs on any OS and any platform. The lobster way.
factory (DROID)		More enterprise-shaped: missions, operators, and team workflows for when agent work starts looking operational.
Codex		OpenAI's lightweight coding agent that runs in the terminal. The one I started with and still reach for when I want something simple and fast.
Claw-code		The fastest-growing coding agent harness I have seen. Built in Rust, feels snappy, and the community around it is moving fast.
OpenClaude		Open-source coding-agent CLI that runs anywhere and uses anything. Good for when I want to swap providers without friction.
paperclip		The open-source app for managing agents at work. Where my experiments stop being a folder of scripts and start looking like something a team could run.

Harness Customisations

Tool	Signal	Why I keep it around
lazycodex		Drops in pre-built shortcuts and lazy automation so I can stay in flow instead of wiring things up by hand.
oh-my-codex		Adds loops like Autopilot and Ralph to Codex, which makes longer-running execution and verification much more practical.
oh-my-openagent		Adds Sisyphus and orchestration-focused agents, which makes OpenAgent feel much more useful as a workflow shell.

Agent Skills

Tool	Signal	Why I keep it around
agent-skills		Production-grade engineering skills for coding agents. The difference between an agent that writes code and one that ships it to a standard.
superpowers		An agentic skills framework and a real software development methodology. Adds workflow skills for planning, brainstorming, and working with more discipline instead of improvising every step.

Creative & Media

Tool	Signal	Why I keep it around
hyperframes		Write HTML, render video, built for agents. The most convincing case I have seen for letting an agent produce real motion graphics instead of just text.

Utilities

Tool	Signal	Why I keep it around
cc-switch		Gives me one place to switch providers, configs, MCP, and skills across Claude Code, Codex, Gemini CLI, OpenCode, and OpenClaw.
CLIProxyAPI		A clean CLI proxy API for routing requests across providers, which keeps the network layer flexible and pairs well with cc-switch.
rtk		Helps me save tokens by filtering, summarizing, and cleaning noisy terminal output before it reaches model context.
ccusage		Helps me track usage, token burn, and cost across Claude Code and Codex without having to guess.
codex-auth		Makes Codex account switching less annoying when I need to move between different accounts or setups.
OpenSpec		Helps me write specs in a more structured way so intent survives the jump from idea to execution.

For Beginners Passing By

Tool	Signal	Why it belongs here
syta-super-launcher		A tool I made for friends who do not really know how to code but still want to get into AI coding without wrestling with the setup first.

Proxy

Tool	Signal	Why I use it
Hiddify		The simplest one for me when I want something easy to set up, easy to explain, and easy to live with.
Clash Verge Rev		The most advanced one in my setup when I want more control, more tuning, and a fuller desktop surface.
ClashMetaForAndroid		What I use on Android when I want that same proxy layer to carry over cleanly on mobile.
HaGeZi DNS Blocklists		Works great for anti-tracking, and fits well with the anti-fingerprinting layers I already keep in place across devices.

Current Direction

Exploring  agent workflows, orchestration UX, and operator tooling
Building   a sharper sense for which experiments deserve to become products
Watching   the space between specs, agents, and practical business value
Meeting    Lille / London / Paris

Meet Me In

Available in Lille, London, and Paris, while based between Beijing and Shanghai, with occasional stops in New York.

Connect

Principles

Useful beats impressive.
Taste is part of the implementation.
Tooling should improve how people think and move.
Experiments are more interesting when they can become businesses.

Navigation Menu