If you're a vibecoder, i want to make your life a little bit easier to understand all the functionalities as a lot of it is blackboxed (for good reason to be honest)
Essentially: I really want to help people understand their process a bit better, dive deep into their sessions and costs and have a visual for everything.
What it is:
Think of it as a control tower for AI-assisted dev work
you don’t have to spelunk through folders and config files or remember a bunch of terminal rituals. It visualizes and manages the setup layer—claude.md/agents.md/etc, skills, agents, hooks, workflows—while staying provider-agnostic (Claude, Codex, Gemini). You still run the actual tool in your terminal; this just makes the environment + files sane.
Ill try to explain a bit more in depth because i know it can prob get overwhelming because theres a lot of stuff.
Functionalities/Pages:
Console - a way for you to manage your terminals into workspaces and ability to split terminals into panes (having the ability to see multiple terminals all at pnce and be able to name sessions and workspaces so you dont lose track of what they were on a high level working on)
Session - a way for you to really understand llm spendings in a very granular way (per message) whether it be subscription or api and for you to have the breakdown of subagents details
Review - reviewing the sessions that happened and asking questions for it
Routing graph - a place that you can update your skills claudez.md and all the instruction files(agents,skills,claud.md) for all your projects. You can easily view it, understand it and clean it up on the app than the filesystem. This is if you wanted to optimize/clean up ur context window to minimize token usage.
Skills/agents - a place for you to easily be able to track all the agents and skills in one place, modify them easily without having to dig around. You can also track which are from plugins, how much tokens they use up (appox) and whether you need to optimize for tokens because skills and agents if put in your global state automatically go into your context window when you open claude.
Workflows: its really to have a lot of control and be able to reuse the same agents rather than spawn new ones. When you go on plan mode - you dont actually spawn specific models and specific order of operations. You trust the llm to figure it out. With workflows you’ll be able to start with ai generation and then easily update things to get the configuration and workflow that you want. You can then “deploy them” so that you can do slash commands right away in claude.
Marketplace: its like a plugins factory that helps u understand plugins a lot better - have an understanding of what pieces you actually download + token count when you download a plugin or a package and also have ai check whether or not its secure before installing it
All in all its really a lot of information to help you not only make better use of your tokens but also have a finer control of what you want to do.
So check the website out, try running it locally since its all open sourced.
If you're a vibecoder, i want to make your life a little bit easier to understand all the functionalities as a lot of it is blackboxed (for good reason to be honest)
Essentially: I really want to help people understand their process a bit better, dive deep into their sessions and costs and have a visual for everything.
What it is: Think of it as a control tower for AI-assisted dev work
you don’t have to spelunk through folders and config files or remember a bunch of terminal rituals. It visualizes and manages the setup layer—claude.md/agents.md/etc, skills, agents, hooks, workflows—while staying provider-agnostic (Claude, Codex, Gemini). You still run the actual tool in your terminal; this just makes the environment + files sane.
Ill try to explain a bit more in depth because i know it can prob get overwhelming because theres a lot of stuff.
Functionalities/Pages:
Console - a way for you to manage your terminals into workspaces and ability to split terminals into panes (having the ability to see multiple terminals all at pnce and be able to name sessions and workspaces so you dont lose track of what they were on a high level working on)
Session - a way for you to really understand llm spendings in a very granular way (per message) whether it be subscription or api and for you to have the breakdown of subagents details
Review - reviewing the sessions that happened and asking questions for it
Routing graph - a place that you can update your skills claudez.md and all the instruction files(agents,skills,claud.md) for all your projects. You can easily view it, understand it and clean it up on the app than the filesystem. This is if you wanted to optimize/clean up ur context window to minimize token usage.
Skills/agents - a place for you to easily be able to track all the agents and skills in one place, modify them easily without having to dig around. You can also track which are from plugins, how much tokens they use up (appox) and whether you need to optimize for tokens because skills and agents if put in your global state automatically go into your context window when you open claude.
Workflows: its really to have a lot of control and be able to reuse the same agents rather than spawn new ones. When you go on plan mode - you dont actually spawn specific models and specific order of operations. You trust the llm to figure it out. With workflows you’ll be able to start with ai generation and then easily update things to get the configuration and workflow that you want. You can then “deploy them” so that you can do slash commands right away in claude.
Marketplace: its like a plugins factory that helps u understand plugins a lot better - have an understanding of what pieces you actually download + token count when you download a plugin or a package and also have ai check whether or not its secure before installing it
All in all its really a lot of information to help you not only make better use of your tokens but also have a finer control of what you want to do.
So check the website out, try running it locally since its all open sourced.
github: https://github.com/OptimiLabs/velocity/