Skip to content
YoepanPublic

About

10 AI engineers working together as a team on your code, with a live virtual office you can watch in your browser.

Topics

Resources

Stars

12 stars

Watchers

0 watching

Forks

Repository files navigation

TenAgent Logo

10-Agent Engineering Swarm Workspace & 2D Virtual Office
An autonomous, contract-first multi-agent framework to take any software idea from concept to production with zero architectural drift.

Clean Architecture Virtual Office Platform Windows Anti-Slop License

TenAgent Virtual Office Monitor Dashboard


The Problem TenAgent Solves

When a single AI coding agent is asked to architect schemas, write frontend components, build backend endpoints, create unit tests, and configure Docker containers in a single conversation context, its context window degrades rapidly.

As a project grows past a few files, three common failure modes occur:

  1. Architectural Drift: The AI forgets constraints established in earlier prompts, reinventing duplicate database tables or altering column names across endpoints.
  2. Contract Mismatch: Frontend API calls hallucinate parameters that the backend never implemented.
  3. Generic Boilerplate ("AI Slop"): Code gets cluttered with obvious comments (e.g. // fetch user, // render screen) and generic, uninspired styling.

TenAgent solves this by structuring your project into 10 specialized virtual engineering roles. Each agent focuses exclusively on its narrow domain of responsibility, communicating strictly through version-controlled specifications (specs/) and formal API contracts.


Core Architecture & Office Pods

TenAgent organizes the engineering lifecycle into 5 Collaborative Office Pods, mirroring a high-performing engineering organization:

flowchart TD
    classDef podStyle fill:#0f172a,stroke:#334155,stroke-width:1px,rx:8px,color:#94a3b8
    classDef founderStyle fill:#1e3a8a,stroke:#3b82f6,stroke-width:1.5px,rx:10px,color:#f8fafc
    classDef contractStyle fill:#0c2340,stroke:#0284c7,stroke-width:1.5px,rx:6px,color:#7dd3fc,stroke-dasharray:4 4
    classDef agentStyle fill:#1e293b,stroke:#475569,stroke-width:1px,rx:6px,color:#f1f5f9
    classDef gateStyle fill:#2a121d,stroke:#e11d48,stroke-width:1px,rx:6px,color:#fda4af
    classDef prodStyle fill:#062d24,stroke:#059669,stroke-width:1.5px,rx:10px,color:#a7f3d0

    Founder(["Founder / User Prompt"]):::founderStyle

    subgraph Pod1 ["Pod 1: Strategy & Architecture (Phase 1)"]
        direction LR
        PM["01. Product Manager<br/><small>PRD & Scope Boundaries</small>"]:::agentStyle
        Arch["02. System Architect<br/><small>Clean Architecture ADR-001</small>"]:::agentStyle
        PM -->|"Feature Stories"| Arch
    end

    subgraph Pod2 ["Pod 2: Contracts & Foundations (Phase 2)"]
        direction LR
        Design["03. UI/UX Designer<br/><small>Screen Flows & DESIGN.md</small>"]:::agentStyle
        DBA["04. Database Engineer<br/><small>3NF Schema & Data Dict</small>"]:::agentStyle
        Contract["Shared Contract Lock<br/><b>specs/api/openapi.yaml</b> & <b>schema.sql</b>"]:::contractStyle
        Design -->|"Tokens"| Contract
        DBA -->|"Entities"| Contract
    end

    subgraph Pod3 ["Pod 3: Concurrent Engineering (Phase 3)"]
        direction LR
        Backend["05. Backend Engineer<br/><small>Domain Use Cases & REST API</small>"]:::agentStyle
        ClientEng["06. Client Engineer<br/><small>Web/Mobile UI Implementation</small>"]:::agentStyle
    end

    subgraph Pod4 ["Pod 4: Hardening & Security (Phase 4)"]
        direction LR
        QA["07. QA Engineer<br/><small>Regression & Boundary Tests</small>"]:::agentStyle
        Security["08. Security Engineer<br/><small>OWASP Top 10 & Secret Audit</small>"]:::agentStyle
    end

    subgraph Pod5 ["Pod 5: Deployment & Release Gate (Phase 5)"]
        direction LR
        DevOps["09. DevOps Engineer<br/><small>Docker & CI/CD Pipelines</small>"]:::agentStyle
        TechLead["10. Technical Lead<br/><small>6 Quality Gates Auditor</small>"]:::gateStyle
        Production[("Verified Production Release")]:::prodStyle
    end

    Founder ==>|"Vision & Tasks"| Pod1
    Pod1 ==>|"Architecture Lock"| Pod2
    Contract ==>|"Parallel Build (No Blocking)"| Backend
    Contract ==>|"Parallel Build (Mock Endpoints)"| ClientEng
    Backend --> Pod4
    ClientEng --> Pod4
    Pod4 ==>|"Audit Pass"| DevOps
    DevOps --> TechLead
    TechLead -->|"APPROVED"| Production
    TechLead -.->|"REJECTED (Diff Notes)"| Backend

    class Pod1,Pod2,Pod3,Pod4,Pod5 podStyle
Loading

The 10 Engineering Roles

# Role Domain Mandate Primary Deliverables Standard Enforced
01 Product Manager Discovery interviews, scoping, user stories, acceptance criteria specs/prd/*.md Anti-slop copywriting
02 System Architect Tech stack selection, Clean Architecture layer boundaries specs/architecture/*.md Strict domain isolation
03 UI/UX Designer User flows, wireframes, accessible color tokens specs/design/*.md, DESIGN.md WCAG AA contrast, >=44px tap targets
04 Database Engineer 3NF schema normalization, ERDs, centralized Data Dictionary specs/database/schema.sql Zero redundant tables or keys
05 Backend Engineer Domain use cases, entities, repository adapters backend/src/... Adheres strictly to OpenAPI spec
06 Client Engineer Web or Mobile user interfaces built in parallel frontend/ or mobile/ Mocks endpoints against OpenAPI spec
07 QA Engineer Edge-case hunter, automated unit & integration tests tests/, specs/qa/*.md Boundary & regression validation
08 Security Engineer OWASP Top 10 auditing, auth/authz verification specs/security/*.md Zero token or secret leakage
09 DevOps Engineer Multi-stage Docker builds, reproducible CI/CD pipelines Dockerfile, .github/ Zero-drift deployment builds
10 Technical Lead Gatekeeper audit against 6 Quality Gates before merge specs/reviews/*.md Rejects substandard or drifting code

Key Capabilities

1. Virtual Office Monitor (dashboard/)

TenAgent includes a zero-dependency, ultra-lightweight web dashboard running on port 7777:

  • Zero Lag: Built entirely with pure semantic HTML5, CSS3, and responsive SVGs (under 60KB total). No bulky canvas or heavy WebGL dependencies.
  • Live Process Inspection: Click on any desk or roster card to open the Engineer Dossier Modal with persona traits, domain mandates, and deliverables.
  • Micro-Step Synchronizer: Accurately reflects CLI activity, file changes, and sub-second log streams without fake work animations.
  • Strict Port Isolation: Dashboard runs on dedicated port 7777 (with auto-range up to 7827), ensuring zero collision with project development servers (ports 3000, 5173, 4000, 8000).

2. Autonomous Multi-AI Council & Debate (council.py)

Connect external API keys (Google Gemini, Groq, DeepSeek, Anthropic Claude, OpenAI, or local Ollama) in the dashboard modal:

  • Dynamic Model Pool: Automatically detects whichever keys are active in specs/swarm-config.json.
  • Adversarial Pre-Flight Debate: Before major architectural decisions or schema creation, the fastest model (e.g. Groq) acts as the Red Team / Challenger to stress-test the proposal, while a secondary model (e.g. Gemini 2.5 Flash) defends and refines the solution.
  • Live Streamed Debates: Transcripts stream directly to specs/terminal.log and appear live in the Virtual Office monitor.
  • Graceful Fallback: If no external keys are configured, TenAgent functions seamlessly under your native CLI engine (e.g. Antigravity) with zero disruption.

3. Anti-Slop Discipline

TenAgent integrates the .agents/skills/antislop* hygiene guidelines:

  • Minimalist Comments: Never write obvious comments (e.g. // fetch user, // render screen). Code must be self-documenting. Comments are strictly reserved for non-obvious business logic or algorithmic edge cases.
  • Concise Commits: Single-line Conventional Commits (feat(auth): add google oauth provider, fix(db): add unique constraint on users.email).
  • Human-Centered UI: Tactile custom inputs, purposeful contrast, and zero generic cookie-cutter templates.

Quick Start Guide

Prerequisites

  • Environment: Windows 10 / 11 (PowerShell or Command Prompt)
  • Python: Python 3.9+ (no external pip packages required for core dashboard or council)
  • AI Assistant CLI: Any CLI assistant (Google Antigravity, Kiro CLI, Claude Code, Gemini CLI, or any LLM runner)

Step 1: Clone Repository

git clone https://github.com/yoepan/tenagent.git
cd tenagent

Step 2: Launch Virtual Office Dashboard

On Windows, simply double-click:

office.bat

Or run via terminal:

python dashboard/serve.py

Open your browser at http://localhost:7777/dashboard/.

Step 3: Start Building with Your AI Assistant

Open a separate terminal window in the project folder and launch your AI assistant (e.g. agy):

agy

Give your master prompt:

"I want to build an automated store POS system with offline inventory and receipt printing. Start Phase 1 discovery."

Watch your 10 virtual engineers collaborate, debate edge cases, lock contracts, and implement your application live in the Virtual Office Monitor!


Directory Structure

├── .agents/
│   ├── roles/             # Detailed playbooks for all 10 engineering roles
│   └── skills/            # Integrated anti-slop filters (code, UI, copywriting)
├── assets/                # Screenshots and repository banner assets
│   ├── logo.png           # High-resolution project logo
│   └── dashboard-preview.png # Screenshot of the 2D Virtual Office
├── dashboard/             # 2D Animated Virtual Office Simulator
│   ├── index.html         # Responsive SVG office interface & Dossier modal
│   ├── serve.py           # Port-isolated HTTP server & CLI presence detector
│   ├── favicon.png        # Brand icon
│   └── favicon.ico        # Multi-resolution favicon (16x16, 32x32, 48x48)
├── specs/                 # Version-controlled contract directory
│   ├── swarm-status.json  # Real-time state machine for all 10 engineers
│   ├── swarm-config.example.json # Safe configuration template (git-tracked)
│   ├── prd/               # User stories & acceptance criteria
│   ├── architecture/      # Architecture Decision Records (ADRs)
│   ├── design/            # Screen flows, wireframes, and DESIGN.md
│   ├── database/          # 3NF schemas and Data Dictionary
│   ├── api/               # Shared OpenAPI 3.1 contracts
│   ├── qa/                # Test matrices and bug reproduction logs
│   ├── security/          # Threat models and OWASP audit reports
│   └── reviews/           # Technical Lead gatekeeper sign-offs
├── AGENTS.md              # Master operating rules and swarm execution lifecycle
├── council.py             # Dynamic Multi-AI Council & debate runner
├── stream.py              # Sub-second terminal logger and stream runner
├── stream.bat             # 1-click terminal streaming wrapper
├── office.bat             # 1-click Windows Virtual Office launcher
├── .gitignore             # Shields local API keys and execution logs
└── README.md              # Repository documentation

Security & Privacy

TenAgent is built with a security-first configuration architecture:

  • Your active API keys are stored locally in specs/swarm-config.json which is strictly ignored by git via .gitignore.
  • Public clones receive specs/swarm-config.example.json with empty credentials.
  • On the first run, the server automatically initializes your local config without exposing any secret keys to GitHub.

License

This project is licensed under the MIT License. You are free to use, modify, and distribute it for both commercial and open-source software projects.

About

10 AI engineers working together as a team on your code, with a live virtual office you can watch in your browser.

Topics

Resources

Stars

12 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages