Command a team
One voice, a whole team. Below the voice layer, Verbative is a multi-channel Agentic Development Environment — group your Claude Code CLI sessions, run several agents in parallel, and put every idea in front of an automatic design-review panel.
Channels — parallel agents
A channel groups one or more Claude sessions — called agents — that run at the same time. Use a channel per project, feature, or area, and add as many agents inside it as you need; each agent is its own live claude session, shown as a native chat in the Agent Workspace — streamed replies, tool cards, and permission prompts you answer with a click or your voice. In the Verbative panel you pick the active channel and agent, and your dictation and the voice key are routed there — switching focus to another agent is a single click. Add a channel or an agent, rename them, and start fresh, resume, or continue any session right from the panel.
When more than one agent is live, Verbative keeps every voice legible. The panel marks the agent that currently has your voice with a 🎙 your voice badge, and each agent announces its channel and name before it speaks (e.g. “Web App, Checkout flow: …”) so you always know who's talking. Every agent's spoken output flows into the one voice queue, each cue tagged with its agent, and a busy agent's row animates so you can see at a glance which agents are working. Each agent also has its own mute, working-pulse, and attention-beep toggles — silence or tune one agent without touching the others, while the global working-pulse and attention-beep toggles in the Channels header stay the master switches over all of them.
Each channel runs its own Claude agents in parallel. Your voice goes to the selected agent.
Wired up the Stripe PaymentIntent with an idempotency key and the charge path.
All 18 tests pass. Edited src/checkout/intent.ts (+24 −3).
Built the welcome step and added form validation.
All 12 onboarding tests pass.
One window: the Verbative panel on the left, each agent as a native chat on the right — streamed replies, tool cards, and permission prompts you answer with a click or your voice.
Every agent in every channel shares one project memory — a fully on-device memory system that captures decisions and conventions from each turn and recalls exactly the right ones into every prompt. It has its own chapter: Agents that remember →
Advisory Board
A standing design-review panel for every prompt. Turn it on and each prompt is reviewed in parallel by your panel of domain advisors — by default Backend, Frontend, Security & Compliance, QA, Product, and DevOps / SRE — each running on Haiku in its own neural voice. An advisor only speaks when the concern is in its lane (out-of-lane piling-on is forbidden); their findings are handed to your Tech Lead (your normal Claude Code CLI session), which writes the code. The same panel then reviews the result and forces at most one revision — bounded, with no infinite back-and-forth.
The board is yours: activate or deactivate any advisor, or add your own with its own name, focus, and voice (click any advisor to read, read-only, the exact instructions it's given). Toggle it in the Verbative panel or say “advisory board” by voice; “repeat review” replays the last batch. Free includes up to 10 reviews a day; Advanced is unlimited.
- { }Backend
- <>Frontend
- ⛨Security
- ✓✗QA
- ◐Product
- ↻DevOps
Each advisor · Haiku · its own voice
Every prompt fans out to the panel; the Tech Lead synthesizes their advice and writes the code.