Skip to content

teamolimpo/TeamOlimpo

Repository files navigation

TeamOlimpo

Kubernetes for AI agents — a meta-orchestrator that coordinates specialist agents with structured handoffs, SOPs, and quality gates.

MCP-native OpenCode Python 3.12+ MIT License Status: Alpha Handoffs

Star the repo if you find this useful — it helps others discover it too.


The Problem

CrewAI, LangGraph, and AutoGen solve multi-agent prototyping fast. But once you have 5-10 agents, the real problem emerges: how do they hand work to each other reliably? How do you audit a 7-agent pipeline? How do you know which agent produced what, with what confidence?

Existing tools give you agents. They don't give you an operating system to coordinate them.

TeamOlimpo is that OS.


Try It (2 minutes)

git clone https://github.com/teamolimpo/TeamOlimpo.git
cd TeamOlimpo
uv sync
opencode .

Then ask Poros (the orchestrator):

"Research async Python patterns and write a technical summary."

Poros will:

  1. Route to Proteo (research) → produces structured handoff
  2. Route handoff to Hermione (technical writing) → produces document
  3. Return the result

Every step writes an auditable handoff file. No output is delivered without one.


What Makes This Different

The Handoff Protocol

No other multi-agent system has this. Every time an agent finishes work, it writes a structured handoff before returning control. Status, priority, deviation reports, quality scores — all in one file.

Handoff { st: done | fail | hold, devi: { cause, corrective_action }, quality_score: 1-5, refs: [...] }

No handoff = task incomplete. Enforced at the agent level.

👉 Full handoff protocol spec →

SOP-Driven Architecture

Standard operating procedures define how work gets done, not just what:

  • cb870dc6 — Handoff protocol
  • 900191a0 — Agent creation standards
  • 5f609a1a — Routing and delegation

SOPs are versioned and referenced, never duplicated.

MCP-Native from Day One

All tools are MCP servers — not REST wrappers, not SDK abstractions. Frameworks like LangGraph and CrewAI are adding MCP support. TeamOlimpo started with it.

Quality Gates

  • Efesto (Python dev): ruff check → ruff format → mypy → pytest before delivery
  • Hermione (writer): confidence markers, gap declarations
  • Proteo (researcher): minimum 2-3 independent sources per key claim

The Team (11 Agents)

Agent Role
Poros Orchestrator — routes, tracks, synthesizes
Atena Agent designer — builds the agents that build
Proteo Multi-source domain researcher
Pythagoras Academic researcher (peer-reviewed sources)
Efesto Python developer — production-ready code
Hermione Technical writer — structured documentation
Euterpe Italian essay writer
Clio Vault archivist — PDF, cataloging, verification
Dike Risk analyst (Emerson DeltaV)
Metis Strategic thinking partner
Eunomia Contextual analyst (email vault)

Quick Comparison

Feature LangGraph CrewAI AutoGen TeamOlimpo
Category State machine Role-based Conversation Meta-orchestrator
Handoff protocol ✅ Mandatory + auditable
SOP-driven ✅ Versioned SOPs
MCP-native 🛠 Adding 🛠 Adding 🛠 Adding ✅ Day one
Quality gates ✅ Per-agent

TeamOlimpo is not a framework for building agents — it coordinates them after you have them.


Project Structure

TeamOlimpo/
├── .opencode/agents/     # Agent prompt files
├── Team/
│   ├── Members/          # Agent identity profiles
│   └── SOPs/             # Standard operating procedures
├── tools/                # MCP servers (Python)
│   ├── synapsis/         # Search, task, handoff, session
│   ├── executor/         # Shell execution with Token Juice
│   └── email_processor/  # Email vault processing
└── Library/              # Private data (gitignored)

Status

Alpha — 777+ handoffs executed, 11 agents operational, core protocol solid.

✅ Works Now

  • Poros orchestration with IntentGate routing (17 categories)
  • Structured handoff protocol (11 types, 4 statuses, deviation reporting)
  • Synapsis tool suite (search, task, session, handoff, admin)
  • All 11 agent profiles with SOP enforcement
  • Quality gates for code and documentation

🔜 Coming

  • AutoGen-to-TeamOlimpo migration tooling
  • Visual handoff graph (auto-generated Mermaid)
  • Pluggable agent SDK adapters
  • Runtime quality analytics
  • PyPI package

Contributing

  • No handoff, no merge — every PR must include handoff specs
  • SOPs before code — new workflows start with an SOP proposal
  • Quality gates — code must pass ruff, mypy, pytest

Start with Team/SOPs/ — read the handoff protocol (cb870dc6) and the agent methodology (900191a0).


Star on GitHub — it helps the project grow
📖 Read on Substack — full story behind the system

MIT License — LICENSE

About

The orchestration layer for AI agent teams — MCP-native orchestrator with unified memory, IntentGate routing, and structured handoff protocol.

Topics

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages