Getting Started
The fastest way to get your team onto the platform is the Logos Agent — a GitHub Copilot coding agent that handles the full onboarding conversation and opens a pull request with every change. No local setup, no YAML to write by hand.
👋 Hi! I'm the Logos Agent. I help manage everything on the osinfra.io platform — teams, members, repositories, environments, and more.
Give me just a moment while I look you up…
It looks like you're not part of any team yet — welcome! Would you like to join an existing team or onboard a new team from scratch?
Based on your team key I'd suggest Velocity as the display name — it appears in Google Cloud Platform, GitHub, and Datadog. Does that work?
What email address should be the Datadog team admin?
Here's what I'll create:
- 📁 GCP folder hierarchy —
st-velocity(sandbox / non-production / production) - 👥 Google Identity groups — admin, reader, writer
- 🐙 GitHub parent team + 4 child teams (sandbox-approvers, non-production-approvers, production-approvers, repository-administrators)
- 🐶 Datadog team — Velocity
Ready to open the pull request on pt-logos?
The Logos Agent
The Logos Agent manages everything a stream-aligned team needs on the platform: GCP folder hierarchy, identity groups, GitHub teams, Datadog team, repositories, environments, and feature flags. It reads the current state from pt-logos and opens a reviewed pull request for every change — nothing is applied directly.
How to invoke it
Open GitHub Copilot and ask it to use the Logos Agent from the pt-logos repository:
Use the Logos Agent in osinfra-io/pt-logos to onboard my team.
The agent will look up your identity, check whether you already exist on the platform, and walk you through the rest.
Clone the repo and run the Copilot CLI from within it — then type /agent and select Logos Agent from the menu:
git clone https://github.com/osinfra-io/pt-logos
cd pt-logos
gh copilot
The agent opens pull requests and reads repository state using the GitHub MCP server. It must be enabled with write toolsets — read-only MCP will allow the agent to inspect state but it will not be able to create branches or open pull requests.
What it can do
| Task | What gets created |
|---|---|
| Onboard a new team | GCP folder hierarchy (sb/nonprod/prod), Google Identity groups, GitHub parent + child teams, Datadog team |
| Add or remove a member | GitHub team membership, Google Identity group membership, Datadog team membership |
| Add a repository | GitHub repo registered in Logos with description, topics, branch protection, and optional feature flags |
| Add a GitHub environment | Deployment environment on a repo with reviewer teams and branch policies |
| Enable a feature flag | Workflows service account, OpenTofu state bucket, Datadog webhook, WIF binding |
| Add a GCP project | Additional Google Cloud project for the team |
| Add a GKE cluster location | Zone-pinned or regional cluster location registered in the team's Kubernetes configuration |
What happens after the PR merges
Logos runs OpenTofu on merge — the GCP folder hierarchy, identity groups, and GitHub structure are created automatically. Corpus and Pneuma pick up the new team data on their next deployment, provisioning GCP projects and Kubernetes namespaces.