Voice is the interface
Speak naturally
No menu maze. No prompt juggling. Just hold, talk, and go.
DotAgents gives you a team of AI specialists with memory, skills, and tools — controlled by your voice. Download the app, speak naturally, and get results without babysitting servers, prompts, or runaway automations.
One dot. Every agent.
Hold to speak. Release to act. Watch agents work with real tools, persistent memory, and confirmation before destructive actions.
$ hold Ctrl+Alt
You: “review the latest PR and summarize what changed”
dotagents
├── listens
├── routes the task to the right agent
├── calls tools with confirmation when needed
└── returns a result you can use immediately
DotAgents is built for real daily use, not just demos. Give a task by voice, let the right agent handle it, and stay in control the whole time.
Voice is the interface
No menu maze. No prompt juggling. Just hold, talk, and go.
Your agents remember
Skills, memory, and persistent preferences carry forward across sessions.
It asks before it acts
Confirmation flows and safe defaults keep capable agents useful without becoming reckless.
DotAgents brings three things together in one system: a voice-first desktop app, an open protocol, and a portable way to package what works.
DotAgents bundles can package agents, skills, MCP setup, and other non-secret configuration into a shareable .dotagents artifact. Safe defaults keep memories and repeat tasks opt-in for public sharing.
$ dotagents://install?
bundle=https://dotagentsprotocol.com/api/bundles/research-stack.dotagents
bundle.manifest.publicMetadata
├── summary: "A shareable agent bundle for the Hub"
├── author: AJ (@techfren)
├── tags: productivity, agents
└── compatibility: macOS desktop
desktop flow
└── open .dotagents → preview contents → resolve conflicts → install
DotAgents is the best way to use the .agents standard — but not the only place it works. Skills and agent configuration can move across tools that adopt the same open file-based convention.
.agents/
├── agents.md # instructions
├── mcp.json # tool configuration
├── models.json # model presets
├── skills/ # reusable capabilities
├── agents/ # specialist agent profiles
├── tasks/ # repeat tasks
└── memories/ # persistent context
Open-ended autonomy sounds exciting until it becomes expensive, unreliable, and hard to control. DotAgents is built for daily use: observable, deliberate, and safe by default.
The product story
Where the difference shows up
DotAgents is a voice-first desktop app for working with AI agents, built on the open .agents standard for portable skills, memory, and configuration.
No. DotAgents is designed as an orchestrator for specialist agents. Voice is the main interface, and the system is built around reusable skills, memory, tools, and delegation.
DotAgents is the AI agent you download, not deploy. It avoids the constant setup, maintenance, and runaway behavior that often come with always-on, self-hosted agent systems.
Public bundle export strips secrets automatically, and sharing defaults are conservative. Memories and repeat tasks are opt-in for public export.
DotAgents routes it into a preview and conflict-aware import flow so you can inspect what’s being installed before applying changes.
Final CTA
Download DotAgents, give it a task, and see what AI agents feel like when they’re built to be used.