Back to Devoured - June 01, 2026
DEVOURED
ECC (GitHub Repo)

ECC (GitHub Repo)

AI agentsdeveloper-toolsmulti-agent GitHub
The ECC (Enterprise Command Center) agent workflow system has reached v2.0.0-rc.1, offering a comprehensive, cross-harness framework for agentic work with skills, memory, and security features.
What: ECC, with over 182,000 GitHub stars and 170+ contributors, is an operator system for agentic work that supports multiple AI agent harnesses like Codex, Claude Code, Cursor, and Gemini. Version 2.0.0-rc.1 introduces the Hermes operator story, enhanced desktop dashboard, and Rust control-plane prototype.
Why it matters: ECC's focus on a "harness-native operator system" across diverse AI agent platforms (like Claude Code, Cursor, OpenCode) highlights the emerging need for standardized, portable, and production-ready frameworks that can manage and orchestrate agents regardless of the underlying AI model or IDE.
Takeaway: Explore ECC's guides (Shorthand, Longform, Security) to understand how to build and deploy complex, production-ready AI agents across various development environments.
Deep dive
  • ECC is a comprehensive system for multi-harness agent workflows, built from real-world engineering use cases over 10+ months.
  • It includes skills, instincts, memory optimization, continuous learning, and security scanning.
  • ECC works across numerous AI agent platforms including Codex, Claude Code, Cursor, OpenCode, Gemini, Zed, and GitHub Copilot.
  • Version 2.0.0-rc.1, released April 2026, introduces the public Hermes operator story and a Tkinter-based desktop dashboard (ecc_dashboard.py).
  • The system now includes 63 agents, 249 skills, and 79 legacy command shims, with selective installation based on manifests.
  • A Rust control-plane prototype (ecc2/) is available in alpha, offering commands like dashboard, start, sessions, and status.
  • ECC Pro is a hosted GitHub App for private repositories, funding the open-source development.
  • Key features include token optimization, memory persistence with automatic context saving, continuous learning via auto-extracted patterns, and verification loops.
  • The system includes a security guide covering attack vectors, sandboxing, sanitization, and AgentShield integration.
  • Cross-platform support for Windows, macOS, and Linux, with hooks and scripts rewritten in Node.js for compatibility.
  • Multi-model commands like /multi-plan require an additional ccg-workflow runtime installation.
Decoder
  • Agent Harness: An environment or platform (e.g., Claude Code, Cursor, Gemini) that hosts and runs AI agents, providing them access to tools and context.
  • Skills: Pre-defined workflows or domain knowledge modules that an AI agent can utilize to perform specific tasks.
  • Instincts: Patterns or behaviors learned by an agent, potentially through continuous learning, influencing its future actions.
  • MCP: Refers to a system for pulling outside data, in this context, similar to how NotebookLM's 'Connectors' feature works.
Original article

Full article content is not available for inline reading.

Read the original article →

Explore all 68 articles in today's digest

Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi