| Who runs the work | VestraYou name your chief and tell it what you need in a sentence. It asks what a good manager would, then writes a teammate card: the job, goals you can check, the apps it uses, its first routines, its engine and what it must ask you about first. You read it, change anything and say yes. From then on your chief hands out the work, makes sure every result is checked before it reaches you, and brings you only real decisions. Everyone on your team gets a chief of their own, and chiefs ask each other for answers, so nobody chases anybody.[9][10][11] | CodexYou run Codex yourself, in the ChatGPT desktop app, the CLI, the IDE extension or Codex Cloud. Ask it to spawn subagents and it runs specialized agents in parallel, then the main thread collects their results; local clients let you define custom agents with their own models and instructions, and at the Ultra level ChatGPT can delegate on its own. Every worker is a Codex thread started from one person’s chat.[12][13] |
|---|
| See your whole team | VestraYour office is a map you can walk around, like a game. Every person, chief and teammate has a desk, and each teammate’s tag shows the engine it runs on, so you can see at a glance that code is written on Claude Code and reviewed on Codex. A Right now feed says in one line what each one is doing, live, and an orange tag shows who is waiting on whom. Tap any desk to see its work, watch its computer live or take over. Nothing moves on the map unless real work is happening.[1][14] | CodexThe desktop app keeps your chats and projects in view so you can run several in parallel, and each subagent gets a thread you can open. On iPhone, Codex lists your tasks in the cloud and on connected computers and shows when your input is needed. Both show your own work.[13][2] |
|---|
| Plans | VestraLite $19 a month for up to 3 teammates and 2,000 credits. Pro $97 for 20 teammates, 10,000 credits and phone calls. Max $197 for 50 teammates and 20,000 credits. Ultra $497 for 150 teammates and 50,000 credits. Billed yearly, they are $17, $87, $177 and $447 a month. One credit is one cent, the same on every engine and for model time, the computer, searches and calls, and every plan includes your chief and the whole office.[3][14] | CodexCodex comes with ChatGPT plans. Free ($0) and Go ($8) get GPT-6 Luna in the desktop app, subject to rollout. Plus is $20 a month and Pro comes at $100, $200 or $500. Business is $20 a user a month billed yearly, or $25 monthly, for 2 or more users, and Enterprise and Edu are priced by sales. You can also pay API rates with an API key, without cloud features. Plus and Business usage resets in five-hour windows, weekly limits may apply, and credits extend it.[4] |
|---|
| Getting started | VestraSign up directly, on any plan, with nothing to install. Every new office starts with 1,000 free credits, enough to hand over real work before you pick a plan. Your chief suggests your first teammates and an engine for each, you connect a few tools, and work starts the same day. Your team’s memory fills from your tools’ history on day one, so it already knows your customers.[3][15][16] | CodexInstall the ChatGPT desktop app for macOS, Windows or Linux, or the CLI with one command (a script, npm or Homebrew), then sign in with ChatGPT or an API key and open a project folder. A free plan is available.[17][18] |
|---|
| Where you use it | VestraTalk to your chief in the Vestra app on the web and the desktop app for Mac, in Slack (by direct message or a mention in any channel), in Telegram, or by email: forward anything, or copy it in on a thread. Call its number from the car, or walk up to its desk in the office and just talk. It’s the same chief everywhere, so a conversation started in Slack can finish in the app, and replying “1” takes the suggested option on any decision.[5][9] | CodexIn the ChatGPT desktop app for macOS, Windows and Linux, on the web, in the terminal, in VS Code and compatible editors through the IDE extension, and in the ChatGPT app for iPhone. Admins can set up @ChatGPT in Slack and Microsoft Teams; delegating Codex Cloud work from Slack needs separate setup.[17][2][19] |
|---|
| Calls and email | VestraYour chief has its own phone number and email address, so it calls and writes from its own line and inbox, not yours. It writes first, replies and follows up until there’s an answer. Before a call it reads what the team knows about the person, and afterwards the notes and next steps go where they belong. The first call or email to anyone new waits for your yes. Phone calls start on Pro.[5][3][20] | CodexCodex’s docs don’t describe phone calls or an email address of its own. Plugins such as Gmail work through an account you or your workspace connect.[7] |
|---|
| App connections | VestraMore than 2,000 apps connect in a click, including Gmail, Slack, HubSpot, Notion, Shopify and Stripe, and engineers’ teammates work in GitHub and Linear. Connect an app once for the whole team, then choose which teammates may use it. Logins stay sealed: agents use them without ever seeing the password, and you can take access back in one tap. Where there’s no connector, a teammate uses the website on its own computer.[20][6] | CodexPlugins bundle skills and MCP servers from one directory shared with ChatGPT, such as Gmail and Google Drive, and you can connect any MCP server. Codex works with GitHub and GitLab repositories. Plugins work in the desktop app and CLI, not the IDE extension, and Business workspaces can share company-managed connections.[7][21] |
|---|
| Model and engine choice | VestraMore than a model: pick the engine that does each teammate’s work, Vestra’s own, Codex or Claude Code. If your engineers already work with Codex, their teammates can run on it too. Your chief suggests one and says why, or picks job by job. Mix them in one team, so code written on Claude Code is reviewed on Codex with fresh eyes. Switch any time without losing the teammate’s job, skills, memory or kit, and every engine works under the same approvals, limits and receipts.[14][5] | CodexRuns on OpenAI models, GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna, with a reasoning level you choose, and each subagent can use its own model. Local clients can also point at other providers, such as Ollama, LM Studio, Amazon Bedrock or Google Cloud, through their config file. The agent doing the work is always Codex.[22][23][12] |
|---|
| Its own cloud computer | VestraYour team works on a cloud computer with a real browser, files and a terminal. It uses any website the way you do, including ones with no connector, like supplier portals, booking pages and government forms. Watch it live, take over with one click when a site needs you, and hand it back to carry on where you left off. It has 8 GB on Lite and Pro, 16 GB on Max and 32 GB on Ultra, and agents can also run in your own cloud or on your premises.[20][3][14] | CodexRuns on your own machine in an OS-enforced sandbox, or in Codex Cloud, where each task gets its own workspace and keeps going while your computer sleeps. The desktop app has a built-in browser, and Computer Use can operate other apps on macOS and Windows.[24][25][26] |
|---|
| Approvals | VestraSpending over your limit, contacting someone new and deleting anything always come to you, on every engine, and no setting turns that off. Everything else you tune teammate by teammate and task by task: ask first, suggest, or act and tell you afterwards. New teammates start by asking, and get more freedom only when you move the setting. Each decision arrives in one line with the facts and a suggestion, so you can answer from your lock screen.[6][9] | CodexA sandbox sets what Codex can touch, with network access off by default, and an approval policy sets when it stops to ask. Ask for approval is the suggested mode; Approve for me sends requests to automatic review instead of you. You can approve requested commands from your phone.[27][26][2] |
|---|
| Spending limits and receipts | VestraCap any single task and the whole day, up to $100 a day. Anything unusually big, over 50 credits by default, waits for your okay with the cost shown. When a cap is reached, work stops at a sensible point and says what’s left. Every receipt splits the cost into model time, actions and computer time, and shows which teammate spent it on which job and engine. Unused credits roll over, and you can top up any time.[6][3] | CodexA usage dashboard, and /status in the CLI, show what’s left of your plan. Business and Enterprise admins can set monthly usage limits per user or group and a workspace overage limit; usage alerts notify but don’t stop spending. The docs don’t describe a per-task cap.[4][8] |
|---|
| Memory | VestraOne shared wiki for the whole team, people and agents on every engine: your people, customers, decisions, numbers and what went wrong last time, with a source for every line. It fills from your tools’ history on day one, then from every task, email and correction. Open any line’s source or edit any page, and the fix reaches every teammate’s next piece of work: correct one teammate once, and they all learn. Export it all as Markdown any time.[16][5] | CodexAGENTS.md files hold instructions you write, layered from your Codex home down to the project, and a project file can be checked into the repo. Local Codex memories are off by default and stored on your machine as generated files you aren’t meant to edit by hand; ChatGPT on the web keeps a separate memory.[28][29] |
|---|
| Schedules and proactive work | VestraEach teammate owns a goal and does three kinds of work: one-off asks, routines that run on a schedule (a market note every Monday at 9) or the moment something happens (a rival changing its prices), and the goal it holds for months. When it spots something worth doing, it brings you a proposal to approve, and on quiet days it says nothing. Your chief sends a morning brief at the time and place you choose.[10][9][10] | CodexScheduled tasks run from ChatGPT on the web or the desktop app, which must stay on for local projects. On eligible plans, tasks on the web and mobile can start from Gmail, Slack or GitHub events. The CLI and IDE extension can’t manage them. On Enterprise, Team Tasks run in the cloud through a team’s service account.[30][31] |
|---|
| Skills | VestraReady-made skills for Sales, Bookkeeping, Content, Customer care and Research, added to any teammate in a click. Custom skills are built around your own playbooks, data and standards, like your code style or release checklist, usually during onboarding, and read like a short plain-language checklist you can edit when your business changes.[10][14] | CodexSkills package instructions, resources and scripts on the open agent skills standard, and plugins share them with others through the directory used by ChatGPT and Codex.[32] |
|---|
| Research | VestraHire a Research teammate in a sentence to watch markets, rivals and people, with sources. Every fact is checked against its source before it counts as done, findings land in your memory wiki with a source for every line, and the rest of your team turns them into the outreach, the deck or the price change.[10][16][11] | CodexRuns live web search from the CLI, uses the desktop app’s browser to gather current information, and splits large investigations across subagents, such as a document of millions of tokens, returning distilled takeaways.[18][33][12] |
|---|