scion

got my PR merged into googlecloudplatform/scion added mocked cloud run lifecycle tests so the runtime can be tested without hitting gcp scion#1453 genuinely forgot to post about this 😳 (my first merged pr btw)

Scion GitHub Overview: How Isolated Agents Scale Beyond Shared Workspaces scion by googlecloudplatform: googlecloudplatform/scion Scion is an open-source orchestration testbed for teams of AI coding agents. This overview explains why parallel agents need distributed-systems infrastructure: the CLI remains a client, the Hub owns identity and state, and a Runtime Broker provisions isolated containers through an OCI-compatible runtime. The architecture separates control-plane decisions from execution while dedicated Git worktrees keep concurrent agents from fighting over one checkout. The discussion also covers Podman, Docker, and the macOS Apple Container fallback, the Git 2.48.0+ requirement for relative-path worktrees, and the Hub's secret-injection model. As-needed credentials preserve least privilege; forcing credentials into every container expands the attack surface. Agent states move through created, running, suspended, error, and deleted, with deletion presented as a destructive teardown. OpenTelemetry forwarders provide structured visibility into sessions, shell commands, and tool use. The same control-plane/execution split supports a single-node VM, Cloud Run identity-aware access, and GKE Autopilot deployment. The final section keeps expectations grounded: Scion is actively developed and pre-1.0, schemas may change, Kubernetes integration has friction, and harness support is uneven. This is an architecture overview, not an installation walkthrough or a claim of successful deployment. TimeStamps: 0:00 Why parallel agents need distributed infrastructure 0:32 Scion's experimental architecture and isolation 1:30 CLI, Hub, and Runtime Broker boundaries 2:44 Container runtimes and network requirements 3:47 Git worktrees for conflict-free concurrency 4:48 Secret injection and least privilege 5:47 Agent lifecycle and destructive teardown 6:42 OpenTelemetry visibility and cloud deployment 8:01 Pre-1.0 limits and harness support 8:40 Scion's infrastructure-first blueprint 🧩 Multi-agent isolation and control-plane boundaries 🗂️ Git worktrees, secrets, and lifecycle state 📡 Telemetry, Cloud Run, and GKE deployment topologies ⚠️ Pre-1.0 limitations and harness compatibility For platform engineers evaluating multi-agent coding infrastructure, Scion’s key tradeoff is clear: stronger isolation and observability require a Hub, a container runtime, and disciplined Git worktree management. Treat the repository as an evolving testbed, verify current harness support and deployment behavior, and weigh its distributed architecture against simpler local workflows. #Scion #AIAgents #GitWorktrees

@ibuildthecloud @ptone of GoogleCloudPlatform/scion fame

Scion: An open foundation for agent collaboration (1.9k stars) What it does All text below is quoted directly from the project README. >An open foundation for agent collaboration. Key Features: - Your choice of agents: Mix supported harnesses and models across a team. Extend agent setup through harness configuration bundles, with provisioning and authentication defined outside Scion's core. - Teams that adapt: Agents can delegate through the CLI, message collaborators, and receive notifications when work finishes or needs attention. Define the collaboration pattern in templates, skills, and instructions. - Workspaces with deliberate sharing: Run agents in containers with their own configuration and credentials. Choose shared directories, Git worktrees, or separate clones according to the project's coordination needs. - People in the conversation: Follow work in native web chat or connect Telegram, Discord, Slack, Google Chat, and Microsoft Teams. Attach to supported interactive harness sessions when you want to work directly with an agent. - Reusable expertise: Package instructions and tools in templates and versioned skills. Publish, discover, and resolve skills through the Skill Bank and connected registries. - Local to hosted execution: Use Docker, Podman, Apple Container, Kubernetes, or Cloud Run. Choose Workstation, single-node hosted, or HA hosted operation, with a managed-agent API path for supported providers. >brew install homebrew-scion/scion/scion Repository GoogleCloudPlatform/scion Repository stats |Metric|Value| |:-|:-| |Stars|1,937| |Forks|287| |Last release|None| |Languages|Go 80.7%, TypeScript 14.1%, Shell 2.3%| |Contributors|38| |AI contributor accounts (Claude / Codex / Cursor / Copilot)|claude: 11, cursoragent: 1| |AI co-authored commits|Claude: 10, Cursor: 1| Main contributor |Field|Value| |:-|:-| |Name|Preston Holmes| |GitHub|@ptone| |Commits to this repo|5,824 (top contributor)| |Company / Location|Google Inc. / Santa Barbara, CA| |Followers|333|

@AnthropicAI Not to be confused with scion from Google: GoogleCloudPlatform/scion