VNAGENTChat · Closed beta

Closed beta · invites rolling out

One chat.A whole team.

VNAgent Chat is where you work with AI agents. Pick the right model for each task, hand big jobs to a lead and its specialist team, and keep every project's memory in one place — saved only when you approve it.

Already invited?

Vietnamese & EnglishClaude · GPT · Grok · DeepSeekOnly beta-access emails

ChatTeamSkills Claude Opus 5.5

↑

What's inside

Four things you'll use every day. No setup, no prompt engineering — just start a chat.

01 — Maestro

Any model

Switch the model behind a chat in one tap — Claude, GPT, Grok or DeepSeek — and set how deeply it reasons.

Claude Fable 5.1Claude Opus 5.5Claude Sonnet 5.5GPT-6.1 SolGPT-6 AstraGrok 4.7DeepSeek V4.1 Flash

02 — Team

A lead + a team

For bigger jobs, a lead model plans the work, hands tasks to specialist members, then reviews and merges what they return.

03 — Skills

A skill library

Reusable skills your agents can call on — so repeat work gets done the same, proven way every time.

researchpdf reportimagespreadsheetemail draftweb fetchsummarize

04 — Projects

Memory you approve

Group chats into projects. When a project learns something, you review it first — nothing is saved without your OK.

Get in

VNAgent Chat is in closed beta. Here's how access works.

Join the waitlist

Leave your email. We only write to you about your beta access.

Get your invite

When a spot opens, your Google account is added to the beta.

Log in & chat

VNAGENTYour AI assistant

Your projects

Recent chats

0 selected

Session tokens0 / 0 · 0%

Compaction—

Maestro · — tok/sHarness · — tok/s

Maestro

Harness

Claude usage

Pending

Account usage · shared with Claude

ChatGPT

—

Quota remaining

Waiting for quota data.

DeepSeek

—

API credit

Waiting for balance data.

Reasoning

Applies from the next message.

New chat

VNAgent / Team

Your agent team, in one place

Loading Team workflow…

Project

＋ New project

——

Agents by role

0 agent

Project members

—

Loading members…

Add work role

Project documents

—

Loading documents…

Add document (txt, md, pdf, zip, png, jpg, webp — max 8 MB per file)

Only text documents (.txt/.md) are added to the bounded context for workers and reviewers. PDF, ZIP and images are stored only: at this stage the AI does not read them, does not OCR them and does not unzip them. Adding/removing documents does not call a model. Each run keeps an immutable snapshot of the documents; later edits/deletes do not change a run in progress. Removal is a soft operation: the row and hash are kept for pinned history, while removed bytes are reclaimed when the project's storage budget is exceeded (hard cap, never grows without bound).

Expertise & knowledge

—

Loading expertise data…

Expert roles

Knowledge · one-way export (test store)

Technical gates · isolated lab (not merged)

Assigning new work from Team is not supported yet. Tasks run on a local deterministic process — NOT a real LLM. Pause/Resume/Cancel/Redo are authenticated lifecycle actions; all external side effects remain fully blocked.

Agent details

Choose an agent to see its tasks, activity and documents. The Complete status appears only after the result is verified and a reviewer approves it.

OpenClaw · Codex Team

Loading Team…

Overall progress—

Members

0 agent

Tasks

All 0

Loading Team dataVNAgent is reading the lead and teammates from OpenClaw.

Skills

Reusable know-how your assistant can call on. Install from GitHub, use in any chat.

Loading skill library…

Team projects

Your projects

Team code

Recent chats

Approvals

Delete all recent chats?

Move chat to project

Rename project

Sync project names

Review project memory

Activity