[{"data":1,"prerenderedAt":116},["ShallowReactive",2],{"blog-\u002Fblog\u002Fbring-your-own-agent-codex-opencode":3},{"id":4,"title":5,"author":6,"body":7,"cover":100,"description":13,"draft":101,"extension":102,"meta":103,"navigation":104,"path":105,"publishedAt":106,"seo":107,"stem":108,"summary":109,"tags":110,"updatedAt":114,"__hash__":115},"blog\u002Fblog\u002Fbring-your-own-agent-codex-opencode.md","Bring Your Own Agent: Claude, Codex, opencode & More","Veflow team",{"type":8,"value":9,"toc":90},"minimark",[10,14,21,26,29,32,35,39,42,50,53,57,60,63,67,70,73,77,80,83],[11,12,13],"p",{},"The desktop ADE in Veflow was built around a simple idea: give an AI coding agent its own isolated git worktree, a real terminal, a full editor, and a docked chat — then get out of the way. Until now, that docked agent was always Claude Code. As of today, it doesn't have to be.",[11,15,16,20],{},[17,18,19],"strong",{},"Veflow's desktop ADE is now agent-agnostic."," When you open the Chat tab in a worktree, Veflow detects which coding-agent CLIs you have on your PATH and lets you pick. One worktree drives Claude Code. Another runs Codex. A third runs opencode. Switch tabs. Compare work. No configuration required beyond having the CLI installed.",[22,23,25],"h2",{"id":24},"what-bring-your-own-agent-means-in-practice","What \"bring your own agent\" means in practice",[11,27,28],{},"When a worktree has no active session yet, the Chat tab shows a picker — your installed agents, nothing more. Click one and a fresh session opens. The agent has the worktree's directory as its working tree, the terminal is wired up, and the editor stays in sync. It's exactly what the Claude Code flow looked like before, except now you choose.",[11,30,31],{},"The currently detected agents include Claude Code, Codex, opencode, Gemini CLI, GitHub Copilot CLI, Aider, and Amp. The registry is extensible — if a CLI announces itself in a way we can detect, it can be added.",[11,33,34],{},"Each session is its own tab in the Chat panel. You can open multiple sessions — different agents on the same worktree, or the same agent running in parallel across two worktrees — and switch between them instantly. Tabs persist until you close them. Nothing is hidden or merged into a single context.",[22,36,38],{"id":37},"zero-veflow-credits-your-subscription-your-rules","Zero Veflow credits. Your subscription, your rules.",[11,40,41],{},"BYO agents run entirely on your machine, against your own API key or subscription for each CLI. Veflow never proxies the calls, never meters the tokens, never debits a credit. Whatever plan you have with Anthropic, OpenAI, or whichever provider the CLI uses — that's the only billing relationship in play.",[11,43,44,45,49],{},"Live token and usage tracking is built into the Chat panel because we know you want to see it. But the usage ",[46,47,48],"em",{},"belongs to you",", settled directly with the provider.",[11,51,52],{},"This is a permanent property of the desktop ADE, not a trial. Local agents on your machine will always be free in Veflow.",[22,54,56],{"id":55},"how-this-differs-from-managed-agents","How this differs from managed agents",[11,58,59],{},"Veflow also ships managed specialist agents — Atlas, Nova, Orion — that run in the cloud on Veflow's infrastructure. They pick up tasks from your board, work autonomously, open pull requests, and settle the cost in Veflow credits. They're the right tool when you want an agent to work while you're away, or when you need to fan work out across more tasks than you have machines for.",[11,61,62],{},"BYO and managed aren't competing — they're different modes. The desktop ADE is an interactive, heads-down coding environment. You're in the loop; the agent is your copilot. Managed agents handle tasks you hand off. Most teams end up using both, for different things.",[22,64,66],{"id":65},"why-agent-agnostic-matters-now","Why agent-agnostic matters now",[11,68,69],{},"The coding-agent landscape moved fast in the last twelve months. Claude Code, Codex, opencode, Aider, Amp — each has real strengths, real tradeoffs, and devoted users. Codex tends toward brevity. Aider's diff workflow has a following. opencode is adding capabilities quickly. Locking the ADE to one CLI was a position we took early and should have reconsidered sooner.",[11,71,72],{},"More practically: you already have a preference. You've tuned your workflow around a specific agent. The right tool is the one that fits how you think, not the one we happened to ship first. Veflow's job is to provide the environment — worktrees, editor, terminal, task board, PR integration — and let you bring the intelligence you trust.",[22,74,76],{"id":75},"getting-started","Getting started",[11,78,79],{},"Download the Veflow desktop app and open any repository. Create a worktree, open the Chat tab, and the agent picker appears if you have multiple CLIs installed. If you only have one detected, it launches automatically — no picker, no friction.",[11,81,82],{},"Your agents. Your machine. Your subscription. The ADE just makes them productive.",[11,84,85],{},[86,87,89],"a",{"href":88},"\u002Fdesktop","Get the Veflow desktop app →",{"title":91,"searchDepth":92,"depth":92,"links":93},"",3,[94,96,97,98,99],{"id":24,"depth":95,"text":25},2,{"id":37,"depth":95,"text":38},{"id":55,"depth":95,"text":56},{"id":65,"depth":95,"text":66},{"id":75,"depth":95,"text":76},"\u002Fblog\u002Fbring-your-own-agent-codex-opencode.png",false,"md",{},true,"\u002Fblog\u002Fbring-your-own-agent-codex-opencode","2026-07-26",{"title":5,"description":13},"blog\u002Fbring-your-own-agent-codex-opencode","Veflow's desktop ADE is now agent-agnostic — bring your own coding-agent CLI (Claude Code, Codex, opencode, Aider, and more) and run it free, forever, on your own subscription.",[111,112,113],"agents","ade","byo",null,"YA4roNyo3UPv27dRwJAxO3rNe3cf0ZhOOdFD220Tuz8",1785474880756]