Manos

A voice-first AI executive that plans, reasons, and securely executes real-world tasks across apps using GPT-5.6 and Codex.

Comment

Manos – Your AI Chief of Staff

Inspiration

Large Language Models are excellent at answering questions, but answering is only a small part of real work.

Professionals spend most of their day switching between emails, calendars, documents, messaging apps, browsers, and development tools. Even with AI, users still have to manually connect these steps together.

We wanted to build something different.

Instead of creating another chatbot, we built an AI Chief of Staff—a voice-first agent that understands goals, plans multi-step workflows, asks for approval before sensitive actions, and executes real work across multiple tools.

Our vision is simple:

Talk naturally. Let AI handle the execution.

What it does

Manos transforms natural conversations into completed tasks.

Users simply describe what they want, and Manos:

  • Understands the user's intent
  • Breaks complex goals into executable subtasks
  • Determines which tools are required
  • Generates an execution plan
  • Requests approval before external actions
  • Executes the workflow
  • Maintains conversation context throughout the process

Examples include:

  • Managing emails
  • Creating presentations
  • Organizing files
  • Generating documents
  • Executing coding tasks
  • Planning multi-step projects
  • Orchestrating multiple AI tools from a single conversation

Rather than responding with text alone, Manos focuses on completing objectives.

How we built it

Manos is built around GPT-5.6 and Codex as its reasoning and execution engine.

The architecture consists of several coordinated components:

  • Voice Input Layer
  • Intent Understanding
  • Planning Engine
  • Task Decomposition
  • Secure Approval Workflow
  • Tool Router
  • Memory Layer
  • Execution Engine

GPT-5.6 provides high-level reasoning, planning, and decision making, while Codex powers software engineering tasks such as code generation, debugging, and implementation.

The backend coordinates multiple tools and services to complete real-world workflows while maintaining user control through approval checkpoints.

Challenges we ran into

Building an AI agent proved very different from building a chatbot.

Some of the biggest challenges included:

  • Determining when the assistant should execute versus ask for clarification.
  • Preventing accidental execution of sensitive actions.
  • Designing reliable multi-step task planning.
  • Coordinating reasoning across different tools.
  • Creating a natural voice-first experience instead of a traditional chat interface.

Balancing autonomy with safety became one of the most important design decisions throughout the project.

Accomplishments that we're proud of

We're especially proud that Manos goes beyond question answering.

Instead, it acts as an AI coworker capable of understanding objectives and driving them toward completion.

Key achievements include:

  • Voice-first interaction
  • Multi-step autonomous planning
  • Secure approval-based execution
  • Real-world tool orchestration
  • Modular agent architecture
  • GPT-5.6 reasoning combined with Codex execution

We believe this creates a significantly more practical AI experience for everyday professional work.

What we learned

This project reinforced an important lesson:

Large language models alone are not enough to build capable AI agents.

Reliable agents require:

  • Planning
  • Memory
  • Tool orchestration
  • Safety mechanisms
  • Human approval
  • Continuous execution

Building these surrounding systems is just as important as choosing the right model.

What's next for Manos

This prototype is only the beginning.

Our roadmap includes:

  • Mobile-first experience
  • Continuous background execution
  • Enterprise integrations
  • Calendar and productivity automation
  • Multi-agent collaboration
  • Long-term memory
  • Personal knowledge graph
  • Enterprise deployment
  • Team collaboration features

Our long-term vision is to make Manos an AI Chief of Staff that professionals can trust to transform conversations into completed work.

Built With

Share this project:

Updates