Most internal agent experiments die in a notebook because no one wants to own the infrastructure — and the work that could have been automated just keeps piling up.

Every department runs its own inbox triage because nothing ties them together

Coordinating agents across accounting, legal, events, and engineering typically means stitching together three different tools that do not share context. The result is duplicated effort, missed handoffs, and a growing list of things that were supposed to be automated months ago.

One deployment repo now runs agents across your whole org

Y Combinator Open, called quartermaster internally, is a multiplayer agent harness that runs inside Slack and on the web. You point a deployment repo at QM, run qm init with your org slug and a Fly.io or AWS target, and the initialization skill walks you through infrastructure, connectors, web sign-in, and live verification. Input is your existing stack — email, documents, databases, GitHub repos — and output is agents that triage inboxes, open pull requests, monitor CI, and track projects in shared channels.

Platform engineers at startups feel this the most

• Platform engineers at 10-to-500 person companies who are tired of building one-off automations that break when a connector changes
• Legal and accounting ops leads who need scheduled inbox triage with label assignments and reply drafts without buying another SaaS seat
• Engineering managers whose teams lose time switching between GitHub, Slack, and project trackers when a CI run fails

The realistic fit is a startup or mid-sized org with at least one person comfortable with infrastructure — not a solo founder running everything from a laptop.

Open-source agent infrastructure is moving faster than enterprise vendors expected

With multi-agent frameworks from Anthropic, Microsoft, and a growing field of well-funded startups all competing for the same deployment budget, a free MIT-licensed harness from a credible institution changes the build-versus-buy calculation overnight. Orgs that lock in an internal standard now will be the ones defining what enterprise agent architecture looks like in 18 months.

What your team can actually do with this today

• Search internal notes, email, documents, and the web in a single agent query
• Schedule inbox triage that outputs labels and draft replies automatically
• Connect to a GitHub repo and have agents run tests, open PRs, and watch CI
• Track a project inside a shared Slack channel with no manual status updates

MIT-licensed and free to deploy — infrastructure and compute costs are yours to own.

QM requires Postgres and a cloud account, so anyone expecting a one-click install will hit a wall fast.

Teams already deep in a managed platform like Slack’s Workflow Builder or a no-code automation tool may find QM overbuilt for their needs. For orgs that want something deployable without a platform engineer, a hosted agent product will get you to production faster.

Agent infrastructure is leaving hosted platforms and moving into org-owned repos

The shift from renting agent capabilities to owning the harness outright is accelerating, and the tools driving it are now free. We cover tools like this every Friday — subscribe here and we’ll send the best ones straight to you.