Awesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration.
Python agent-harnessagent-memoryagent-orchestrationai-agent-harnessai-agentsawesome-listcontext-engineeringharness-engineeringmcp
Add GoogleCloudPlatform/scion to Task Runners & Orchestration
ai-boost8aecd301f90b
diff --git a/README.md b/README.mdindex 4294981..e1939b2 100644--- a/README.md+++ b/README.md@@ -332,6 +332,7 @@ Harness components organized by the problem they solve, not by vendor.- [Orca](https://github.com/stablyai/orca) — The AI orchestrator for running Codex, Claude Code, OpenCode, and Pi side-by-side in isolated git worktrees. It turns fleet-of-agents execution into a polished desktop IDE with a mobile companion, making parallel agent harness design accessible beyond shell-scripting teams. - [LoopX](https://github.com/huangruiteng/loopx) — Provider-neutral control plane that sits above existing harnesses (Codex, Claude Code, Cursor) and keeps long-horizon objectives, gates, todos, evidence, and quota stable across bounded agent turns. The local-first state kernel makes multi-day work reviewable, restartable, and handoff-safe without replacing the runtime underneath. - [HarnessRouter](https://github.com/HarnessRouter/harnessrouter) — Self-hosted unified API for running Codex, Claude Code, DeepSeek Harness, Pi, and other terminal agents through the open Unified Harness Protocol (UHP), with sessions, streaming, file access, cancellation, and failure handling. Worth including because harness fragmentation is becoming a real operations problem — this is the clearest open-source attempt to make agent runtimes interchangeable behind a single control plane rather than locking teams into one CLI. +- [Scion](https://github.com/GoogleCloudPlatform/scion) — Google's experimental "hypervisor for agents" (April 2026, Apache-2.0): orchestrates Claude Code, Gemini CLI, Codex, and OpenCode as isolated concurrent processes, each with its own container, git worktree, credentials, and network policy. The core harness insight is *isolation over coordination* — let agents run fully autonomous inside their containers and enforce guardrails at the infrastructure layer, so agents that can't touch each other's state can't conflict. The most concrete first-party reference for treating fleet isolation, not prompt-level behavioral rules, as the primary multi-agent reliability mechanism. ### Verification & CI Integration