SUTRA.team – The First OS for Autonomous Agents

TL;DR

TL;DR: SUTRA.team is an operating system for autonomous AI agents — it runs 15 specialist agents as a deliberating council under the Sammā Suit 8-layer security framework, synthesizing their outputs into unified decisions.

Source and Accuracy Notes

• Project page: sutra.team — verified via direct fetch
• Source repository: github.com/OneZeroEight-ai/portable-minds — README verified
• License: MIT (confirmed in README)
• HN launch thread: news.ycombinator.com/item?id=47218056
• Source last checked: 2026-07-16

What Is SUTRA?

SUTRA.team positions itself as “The First OS for Autonomous Agents” — a platform that orchestrates multiple specialized AI agents working together as a council, rather than a single monolithic AI handling everything.

The core product is built around portable agent definitions in a format called PMF (Portable Mind Format). The GitHub repo ships 15 production-ready agents, each modeled after the Noble Eightfold Path — Wisdom Judge, Purpose Auditor, Communicator, Ethics Judge, and so on. The agents deliberate as a council, with the Sutra agent serving as the synthesis layer that reconciles their competing perspectives.

The Sammā Suit 8-layer security framework underpins the whole system, though the specific technical details of each layer would need to be sourced directly from the Sutra documentation.

Agents Included

The 15 agents cover distinct roles:

| Agent | Role | |---|---| | The Wisdom Judge | Strategic Analyst | | The Purpose | Intention Auditor | | The Communicator | Message Strategist | | The Ethics Judge | Ethical Impact Analyst | | The Sustainer | Sustainability Analyst | | The Determined | Execution Strategist | | The Aware | Pattern Analyst | | The Focused | Deep Analyst | | Legal Analyst | Contract & Regulatory Specialist | | Financial Strategist | Valuation & Capital Strategy | | Technical Architect | Systems Design & Engineering | | Market Analyst | Competitive Intelligence | | Risk Assessor | Threat Modeling & Scenario Planning | | Growth Strategist | Go-to-Market & Acquisition | | Sutra | Synthesis — Council Reconciliation |

Setup Workflow

Prerequisites

• Unix/macOS or Windows
• One of: Claude Code, Cursor, GitHub Copilot, or Gemini CLI

Install via cURL (Unix/macOS)

curl -fsSL https://raw.githubusercontent.com/OneZeroEight-ai/portable-minds/main/install.sh | bash

Install via PowerShell (Windows)

irm https://raw.githubusercontent.com/OneZeroEight-ai/portable-minds/main/install.ps1 | iex

Manual Clone

git clone https://github.com/OneZeroEight-ai/portable-minds.git
cd portable-minds
./install.sh

The installer supports four target platforms: Claude Code, Cursor, GitHub Copilot, and Gemini CLI. After installation, each agent is a self-contained JSON file under agents/ that you load directly into your tool.

Deeper Analysis

How the Council Works

Unlike single-agent systems where one LLM handles everything, SUTRA runs agents as specialists that debate and deliberate. Each agent maps to a specific aspect of the Eightfold Path (Right View, Right Intention, Right Speech, etc.) and contributes a perspective before Sutra reconciles them into a unified output.

The PMF (Portable Mind Format) is at the core. Each agent definition is a JSON file with fields for role, skills, memory, and conversations. The memory field enables portable user context that travels with the persona — a notable design choice for multi-session workflows.

Portability as a First-Class Feature

The README emphasizes “one-command install” for four different AI coding tools. Rather than lock users into one platform, SUTRA ships with converters that translate PMF JSON into tool-specific formats (.mdc for Cursor, custom instruction formats for Copilot, GEMINI.md for Gemini CLI).

Converter stability is tiered:

• Stable: Claude Code — primary supported target
• Secondary: Cursor, GitHub Copilot, Gemini CLI — format may change with upstream releases

What’s New in v1.1.0

• memory field — portable user context
• conversations field — cross-device chat history built into the PMF spec
• IceClaw.online uses PMF v1.1.0 natively

Practical Evaluation Checklist

• [ ] Clone and run install.sh on a fresh Unix/machine
• [ ] Verify 15 agent JSON files appear in agents/
• [ ] Load the Wisdom Judge agent into Claude Code and run a test query
• [ ] Load the Sutra synthesis agent and verify it produces a reconciled output from multiple specialist inputs
• [ ] Test the memory field persistence across sessions
• [ ] Verify the Cursor converter produces a valid .mdc file
• [ ] Assess the Sammā Suit security framework documentation depth

Security Notes

The Sammā Suit 8-layer security framework is presented as a core component of SUTRA. Specific layer details, threat models, and how the framework mitigates prompt injection or agent misalignment should be verified directly from the official Sutra documentation before relying on it for high-stakes deployments.

FAQ

Q: Does SUTRA replace my existing AI coding tool? A: No. SUTRA loads into existing tools (Claude Code, Cursor, Copilot, Gemini CLI) as a layer of agent orchestration. It does not replace the underlying tool.

Q: What happens if the specialist agents disagree? A: The Sutra synthesis agent is designed to reconcile disagreements and produce a unified output. The exact reconciliation logic depends on the PMF specification and the specific agent configuration.

Q: Is the PMF format tool-specific or universal? A: Universal at the JSON level — each agents/*.json file is the source of truth. Converters translate PMF into tool-specific formats (.mdc, GEMINI.md, etc.).

Q: Is the Sammā Suit security framework open-source? A: The agents and PMF specification are MIT-licensed on GitHub. The Sammā Suit framework details should be verified from the official site — the exact scope of what is open vs. proprietary is not fully specified in the README.

Conclusion

SUTRA.team takes a distinctive approach to agent orchestration: instead of one AI doing everything, it runs a council of specialized agents that deliberate before synthesizing decisions. The portable mind format (PMF) is a pragmatic bet on multi-platform portability, supporting Claude Code, Cursor, Copilot, and Gemini CLI from a single agent definition.

The Eightfold Path framing is unconventional and gives the agent roles semantic names that hint at their function (Ethics Judge, Sustainer, Risk Assessor). Whether this is a meaningful architectural choice or branding is something the code and documentation will clarify on closer inspection.

For developers who want to structure multi-agent deliberation as part of their AI workflow — rather than relying on a single model — SUTRA provides a ready-made agent council with 15 pre-built roles and a unified synthesis layer. Try the one-command install if you want to explore the council model in practice.

---

This post was written after reviewing the SUTRA.launch HN thread and the portable-minds GitHub repository README directly.