Skip to content

Hermes Ecosystem Discovery - August 2026

A snapshot of notable open-source tools discovered during an ecosystem scan of AI agent infrastructure: MCP servers, agent governance, browser automation, terminal safety, memory management, and cost control.

Top Tools Found

loop-engineering (10,448 ★) - Agent Loop Governance

Loop engineering patterns with a CLI (loop-audit, loop-init, loop-cost) for agent orchestration systems. Helps teams audit and control agent execution loops. MIT license, actively pushed. github.com/cobusgreyling/loop-engineering

solace-agent-mesh (4,956 ★) - Event-Driven Multi-Agent Mesh

Event-driven multi-agent orchestration framework with MCP and A2A support. Enterprise-grade (Solace), Apache-2.0. github.com/SolaceLabs/solace-agent-mesh

loushang (1,166 ★) - AI Agent Harness

Agent harness with multi-model routing, stateful sessions, tool governance, and provider routing (GPT/Claude/DeepSeek/Qwen/Kimi/GLM/MiniMax). Apache-2.0, fast growth (1.1K stars in under 3 months). github.com/zhnt/loushang

devspace (4,345 ★) - Minimal Coding Agent Harness over MCP

Minimal coding agent harness over MCP for ChatGPT, Claude, Hermes, Grok Bot, and OpenClaw. TypeScript, MIT, active since June 2026 (4.3K stars in under 3 months). github.com/Waishnav/devspace

brigade (3,245 ★) - Personal Intelligence, Enterprise-Grade

Crew of AI agents on a real org chart with one shared long-term memory (Tideline); agents delegate, switch models mid-task, and act across 1,000+ apps. Self-hosted, no account, no telemetry, keys never leave the machine. TypeScript, MIT, pushed daily. github.com/spinabot/brigade

BrowserWing (1,400 ★) - Browser as MCP

Turns browser actions into MCP commands, so any MCP client can drive a browser session. github.com/browserwing/browserwing

shellfirm (926 ★) - Terminal Safety

Blocks destructive shell commands before execution by matching them against a risk pattern library. github.com/kaplanelad/shellfirm

cordum (494 ★) - Agent Action Firewall

Enforces policy and human approval before risky tool calls. Useful for production agent deployments. github.com/cordum-io/cordum

Worth Watching

  • OpenViking (27K ★) - Self-evolving context database
  • cossistant (709 ★) - Customer support AI platform
  • tgo (571 ★) - AI customer service with RAG

Scan methodology: GitHub search across 12 categories including MCP servers, agent governance, browser automation, terminal safety, memory management, and cost control.

Contact Terms Privacy Pricing