A multi-agent orchestration framework evolving into a decentralized marketplace for autonomous agent workflows on Ethereum Optimism L2, with enterprise-grade PKI security, capability-based architecture, and flexible payment systems.
Imagine a world where:
You need a complex legal contract drafted, cloud infrastructure deployed, and medical research analyzed - but you don't want to hire three different expensive consultants, wait weeks for delivery, or worry about whether they'll actually do good work.
Team Agent makes this possible by creating a marketplace where specialized AI agents compete to help you, with:
- ✅ Transparent pricing - See exactly what you'll pay, in whatever currency you prefer (dollars, crypto, even university research credits)
- ✅ Verified quality - Every agent has a reputation score based on real performance, enforced by blockchain
- ✅ Permanent records - All work is cryptographically signed and stored permanently, so you can prove who did what
- ✅ Specialization - Instead of one generic AI, you get experts: a Legal AI that knows California employment law, an AWS AI that's deployed thousands of servers, a Research AI that's read every medical journal
- ✅ Community-driven - Bad actors get penalized, good ones get rewarded, all decided transparently by the community
For Researchers:
"I need to analyze 10,000 medical studies on a new treatment, design an experiment, and write a grant proposal."
→ Team Agent finds Literature Review AI + Experiment Design AI + Grant Writing AI, they collaborate on your project, and you pay with your university's research credits.
For Startups:
"I need to deploy a scalable web app on AWS with auto-scaling, a database, and monitoring."
→ AWS Specialist AI generates complete infrastructure-as-code (Terraform), you review it, approve it, and it's deployed. All work is signed and auditable for compliance.
For Legal Teams:
"I need 50 standard employment contracts customized for California law."
→ Legal Specialist AI generates them in minutes, Critic AI reviews for issues, all contracts are cryptographically signed for authenticity.
Right now, AI assistants are isolated - ChatGPT can't talk to Claude, your company's AI can't share knowledge with a university's AI, and there's no way to pay for or trust specialized AI services.
Team Agent changes this by creating:
- A marketplace where specialized AIs advertise their skills
- Blockchain verification so you know the AI actually did the work (not a human pretending)
- Flexible payments - pay in crypto, dollars, research credits, or whatever the community accepts
- Reputation systems - AIs that do good work get more business, bad ones get penalized
- Permanent records - Everything is cryptographically signed and stored on IPFS, so you have proof
Think of it like Uber, but for AI services - with blockchain ensuring nobody can cheat, and the community deciding what's valuable.
Team Agent is transforming from a standalone orchestrator into a decentralized agent marketplace where:
- 🤝 Agents discover each other via Agent2Agent (A2A) protocol
- 🔌 External systems invoke capabilities via Model Context Protocol (MCP)
- ⛓️ Workflows execute on-chain via Ethereum Optimism L2 smart contracts
- 📦 Artifacts stored permanently on IPFS/Filecoin
- 💰 Flexible payments supporting ETH, OP, USDC, custom tokens, and alternative value stores
- 🔬 Research assistant platform for scientists and academic institutions
See Decentralized Marketplace Vision for the complete roadmap.
┌─────────────────────────────────────────────────────────────────────────────┐
│ MISSION INPUT │
└────────────────────────────────┬────────────────────────────────────────────┘
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ ORCHESTRATOR │
│ • PKI Manager (3-tier CA hierarchy: Government, Execution, Logging) │
│ • Agent Manager (Registration & tracking) │
│ • Capability Registry (Dynamic discovery) │
└────────────────────────────────┬────────────────────────────────────────────┘
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ WORKFLOW EXECUTION PIPELINE │
│ │
│ Phase 1: Architecture │
│ ┌────────────────────────────────────────────────────────────────┐ │
│ │ Architect (EXECUTION domain) │ │
│ │ - Analyzes requirements and designs system architecture │ │
│ └────────────────────────────────────────────────────────────────┘ │
│ ▼ │
│ Phase 2: Implementation │
│ ┌────────────────────────────────────────────────────────────────┐ │
│ │ DynamicBuilder + Specialist Selection │ │
│ │ Current specialists: │ │
│ │ • Legal Specialist (contracts, legal documents) │ │
│ │ • AWS Cloud Specialist (Terraform, CloudFormation, boto3) │ │
│ │ • Azure Cloud Specialist (ARM, Terraform, Azure SDK) │ │
│ │ • GCP Cloud Specialist (Deployment Manager, gcloud) │ │
│ │ • OCI Cloud Specialist (OCI CLI, Terraform) │ │
│ └────────────────────────────────────────────────────────────────┘ │
│ ▼ │
│ Phase 3: Review │
│ ┌────────────────────────────────────────────────────────────────┐ │
│ │ Critic (EXECUTION domain) │ │
│ │ - Reviews code/documents, identifies issues, scores quality │ │
│ └────────────────────────────────────────────────────────────────┘ │
│ ▼ │
│ Phase 4: Recording │
│ ┌────────────────────────────────────────────────────────────────┐ │
│ │ Recorder (LOGGING domain) │ │
│ │ - Publishes artifacts, creates audit trail, signs outputs │ │
│ └────────────────────────────────────────────────────────────────┘ │
│ │
│ Optional: Governance (GOVERNMENT domain) │
│ • Pre-build policy checks • Post-review compliance verification │
└─────────────────────────────────────────────────────────────────────────────┘
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ ARTIFACTS & AUDIT TRAIL │
│ • Generated code/documents │
│ • Workflow JSON record │
│ • TuringTape (JSONL append-only log - cryptographically signed) │
└─────────────────────────────────────────────────────────────────────────────┘
Three-tier certificate hierarchy with cryptographic signing on all operations:
Root CA (self-signed, 10-year validity)
├── Government/Control Plane CA (5-year) → Governance agent
├── Execution Plane CA (5-year) → Architect, Builder, Critic
└── Logging/Artifact Plane CA (5-year) → Recorder
All workflow operations are signed:
- ✅ TuringTape entries
- ✅ Agent outputs
- ✅ Artifact manifests
- ✅ Audit logs
Test Coverage: 56/56 tests passing (including 17 PKI tests)
# Python 3.11+
python --version
# Node.js 20+ (for frontend)
node --version# Navigate to main directory
cd swarm2/team-agent
# Install Python dependencies
pip install -r requirements.txt
pip install -e .
# Run tests to verify installation
pytest utils/tests/ -v
# Expected: 56/56 tests passing# Simple demo
python examples/simple_demo.py
# Interactive mode
python examples/interactive_demo.py
# Mission-based execution
python examples/mission_demo.py --simple# Terminal 1: Backend
cd swarm2/team-agent/backend
python app.py
# Terminal 2: Frontend
cd swarm2/team-agent/frontend
npm install
npm run dev
# Open http://localhost:5173ta_base/
├── README.md (this file)
│
└── swarm2/team-agent/ # Main implementation
├── swarms/team_agent/ # Core system
│ ├── roles/ # Architect, Builder, Critic, Recorder, Governance
│ ├── specialists/ # Domain-specific agents (Legal, AWS, Azure, GCP, OCI)
│ ├── capabilities/ # Capability implementations
│ │ ├── cloud/ # Cloud infrastructure (AWS, Azure, GCP, OCI)
│ │ ├── medical/ # Medical documentation
│ │ └── legal/ # Legal document generation
│ ├── crypto/ # PKI, signing, verification, CRL, OCSP
│ ├── mcp/ # Model Context Protocol (MCP) server
│ ├── state/ # TuringTape, HITL coordination
│ ├── orchestrator.py # Main workflow coordinator
│ └── agent_manager.py # Agent registration & tracking
│
├── backend/ # Flask API server
│ ├── app/
│ │ ├── api/ # REST endpoints
│ │ ├── models/ # Database models
│ │ ├── services/ # Business logic
│ │ └── database.py # SQLAlchemy setup
│ └── app.py # Flask application entry
│
├── frontend/ # Vue.js + PrimeVue UI
│ ├── src/
│ │ ├── views/ # Page components
│ │ ├── components/ # Reusable components
│ │ └── router/ # Vue Router setup
│ └── package.json
│
├── docs/ # Comprehensive documentation
│ ├── DECENTRALIZED_MARKETPLACE_VISION.md # Complete architecture plan
│ ├── GITHUB_WORKFLOW.md # PR/issue management guide
│ ├── QUICK_START_GITHUB.md # GitHub CLI quick reference
│ └── PKI_ENHANCEMENTS_COMPLETE.md # PKI documentation
│
├── scripts/ # Utility scripts
│ ├── cleanup_duplicate_agents.py # Database maintenance
│ └── create_prs.sh # Automated PR creation
│
├── examples/ # Demo scripts
├── missions/ # YAML mission definitions
├── utils/tests/ # Test suite (56 tests)
└── output/ # Generated artifacts
| Document | Description | Status |
|---|---|---|
| Decentralized Marketplace Vision | Complete 36-week roadmap for blockchain integration | ✅ Complete |
| GitHub Workflow Guide | PR/issue management, labels, milestones | ✅ Complete |
| Quick Start (GitHub) | GitHub CLI reference for contributors | ✅ Complete |
| CLAUDE.md | Development guide for Claude Code | ✅ Complete |
| PKI Enhancements | PKI infrastructure details | ✅ Complete |
| Component | Status | Details |
|---|---|---|
| Core Agents | ✅ Complete | 4 role agents + 5 specialists |
| PKI Infrastructure | ✅ Complete | 3-tier CA, signing, CRL, OCSP |
| Agent Manager | ✅ Complete | Registration, tracking, trust scores |
| Capability Registry | ✅ Complete | Dynamic discovery, keyword matching |
| Cloud Specialists | ✅ Complete | AWS, Azure, GCP, OCI provisioning |
| Legal Specialist | ✅ Complete | Contract and legal document generation |
| Web Frontend | ✅ Complete | Vue.js + PrimeVue UI |
| Backend API | ✅ Complete | Flask + SQLAlchemy REST API |
| Test Suite | ✅ Complete | 56/56 tests passing |
| Database | ✅ Complete | Agent cards, capabilities, mappings |
| Component | Status | Target |
|---|---|---|
| A2A Protocol | ✅ Complete | .well-known/agent.json endpoint |
| MCP Server | ✅ Complete | HTTP/WebSocket capability invocation |
| Agent Discovery | ✅ Complete | External agent card discovery |
| Phase | Component | Timeline |
|---|---|---|
| Phase 3 | Ethereum Optimism L2 Integration | Weeks 5-20 |
| • Smart contracts (3 contracts + DAO) | ✅ Complete | |
| • IPFS artifact storage | ✅ Complete | |
| • Blockchain client integration | ✅ Complete | |
| Phase 4 | Flexible Payment System | Weeks 21-24 |
| • Multi-token support (ETH, OP, USDC) | Week 21 | |
| • Custom value stores | Week 22 | |
| Phase 5 | Research Assistant Platform | Weeks 25-32 |
| • Research specialists (5 agents) | Weeks 25-28 | |
| • Academic payment models | Weeks 29-32 | |
| Phase 6 | Security & Mainnet Launch | Weeks 33-36 |
| • Smart contract audit | Week 33-34 | |
| • Mainnet deployment | Week 35-36 |
See full roadmap for details.
from swarms.team_agent.orchestrator import Orchestrator
orchestrator = Orchestrator()
result = orchestrator.execute("Create a Python REST API for user management")
# → Architect designs API structure
# → Builder generates FastAPI code
# → Critic reviews code quality
# → Recorder publishes artifactsorchestrator = Orchestrator()
result = orchestrator.execute(
"Deploy a scalable web application on AWS with auto-scaling and RDS database"
)
# → AWS Specialist selected (keyword matching: "AWS")
# → Generates Terraform + CloudFormation + boto3 code
# → Full infrastructure as coderesult = orchestrator.execute("Generate an employment contract for California")
# → Legal Specialist selected
# → Generates comprehensive employment contract
# → Includes state-specific clausesTeam Agent is evolving into a marketplace where:
-
Agent Discovery (Phase 2)
- Agents publish capabilities via A2A protocol
- External systems discover and invoke via MCP server
- Dynamic capability matching and routing
-
Blockchain Execution (Phase 3)
- Workflows execute on Ethereum Optimism L2
- Smart contracts manage workflow state transitions
- IPFS storage for permanent artifact availability
-
Flexible Payments (Phase 4)
- Multi-token support: ETH, OP, USDC, custom tokens
- Alternative value stores (research credits, compute time, "magic jelly beans")
- DAO governance for custom payment methods
-
Research Platform (Phase 5)
- Specialized agents for academic research
- Literature review, data analysis, experiment design
- Grant-funded workflows and open science bounties
-
DAO Governance (Phase 6)
- Reputation staking and slashing
- Community-driven policy enforcement
- Decentralized trust scoring
Read the full vision: Decentralized Marketplace Architecture
We welcome contributions! See our GitHub Workflow Guide for:
- Branch naming conventions
- PR templates and labels
- Commit message format
- Review process
- Milestone planning
- Fork the repository
- Create a feature branch
git checkout -b feat/your-feature-name
- Make changes and commit
git commit -m "feat(component): add new feature" - Push and create PR
git push origin feat/your-feature-name gh pr create --template feature.md
See Quick Start (GitHub) for detailed examples.
cd swarm2/team-agent/backend
pip install -r requirements.txt
python app.py
# API available at http://localhost:5002cd swarm2/team-agent/frontend
npm install
npm run dev
# UI available at http://localhost:5173cd swarm2/team-agent
# All tests
pytest utils/tests/ -v
# Specific test suite
pytest utils/tests/test_pki.py -v
# With coverage
pytest utils/tests/ -v --cov=swarms --cov=utils# Format code
black swarms/ utils/ backend/ --line-length 100
# Sort imports
isort swarms/ utils/ backend/ --profile black
# Type checking
mypy swarms/ utils/ backend/- Total Lines of Code: ~15,000+
- Test Coverage: 56/56 tests passing (100% core functionality)
- Agents: 4 role agents + 5 specialist agents = 9 total
- Capabilities: 6+ registered (Legal, AWS, Azure, GCP, OCI, HRT Guide)
- Documentation: 5 comprehensive guides (2,500+ lines)
- Smart Contracts: 5 planned (WorkflowConductor, CapabilityMarketplace, ReputationDAO, PaymentRouter, CustomValueRegistry)
✅ Multi-Agent Orchestration - 4-phase workflow (Architect → Builder → Critic → Recorder) ✅ Capability-Driven Architecture - Dynamic specialist selection via keyword matching ✅ Enterprise PKI Security - 3-tier CA hierarchy with cryptographic signing ✅ Cloud Infrastructure Specialists - AWS, Azure, GCP, OCI provisioning ✅ Legal Document Generation - Contracts, agreements, compliance documents ✅ Web Dashboard - Modern Vue.js + PrimeVue frontend ✅ Comprehensive Testing - 56 tests covering core functionality and PKI ✅ A2A Protocol - Agent discovery and federation (Phase 2) ✅ MCP Server - External capability invocation (Phase 2) ✅ Blockchain Integration - Optimism L2 smart contracts (Phase 3) ✅ IPFS Storage - Decentralized artifact storage (Phase 3) 🔲 Flexible Payments - Multi-token and custom value stores (Phase 4) 🔲 Research Platform - Academic and scientific agents (Phase 5)
MIT License - See LICENSE for details.
- Ethereum Optimism for L2 blockchain infrastructure
- IPFS/Filecoin for decentralized storage
- OpenAI and Anthropic for LLM technology
- Vue.js and PrimeVue for frontend framework
- Flask and SQLAlchemy for backend architecture
- Documentation: swarm2/team-agent/docs/
- Issues: Use GitHub issue templates (bug reports, feature requests)
- Discussions: GitHub Discussions for questions and ideas
- Contributing: See GitHub Workflow Guide
Team Agent: Building the future of decentralized autonomous agent collaboration.
From standalone orchestrator to decentralized marketplace - one capability at a time.