RBA Fleet
RBA Fleet is the platform this catalog is growing into: a governed, internal AI agent platform that any team at RBA can build on. This section is the home for everything we’re building toward it — the strategic plan, the Runner Registry, the Gateway, and the execution layer.
RBA Skills stays the name of the catalog itself. RBA Fleet is the runtime built
on top of it.
The three layers
Section titled “The three layers”A real agent platform is three layers stacked, with a single governance spine running through all of them. Most internal “AI catalog” efforts only build the first.
| Layer | What it is | RBA today | Read more |
|---|---|---|---|
| 1 — Catalog | Where people find and install tooling (this repo + marketplace) | ✅ Strong | the catalog you’re in |
| 2 — Gateway | The one server every agent goes through to reach internal systems | ⚠️ Seed only (rba-hub-mcp) | Gateway |
| 3 — Fleet | Sandboxed Runners that actually do the work | 🟡 Registry live; runtime not built | Runners · Execution & hosting |
The organizing principle
Section titled “The organizing principle”This repo stays the declarative control plane; the Gateway and the Fleet runtime are runtime services that live outside it. RBA Skills is already the “source-of-truth registry” — git-backed, PR-reviewed, validator-gated, with audit-via-git-history. We lean into that: the repo declares and governs tools and Runners; separate services run them.
No large org runs marketplace.json as its production backbone. Neither will we:
marketplace.json is the client distribution surface, the Gateway is the
integration backbone, and the runners/ directory is the Registry.
Naming
Section titled “Naming”Plain infrastructure vocabulary, so the platform is easy to adopt across the org.
| Term | What it means |
|---|---|
| RBA Fleet | The overall agent platform (all three layers) |
| Runner | A single autonomous agent that does one job |
| the Fleet | The Runners collectively |
| Gateway | The one server every Runner goes through to reach internal systems |
| Registry | The source-of-truth list of every Runner (declared as code in this repo) |
| Sandbox | The isolated environment a Runner executes in |
Where to go next
Section titled “Where to go next”- Platform Roadmap — the full sequenced plan, the workstreams, the risks, and the research it’s grounded in.
- Runners — the Registry: how to declare a Runner and the governance bar it must clear. (Live now.)
- Gateway — Layer 2: the single chokepoint, scoped tokens, and the unified audit log. (Designed, not yet built.)
- Execution & hosting — Layer 3: the Claude Agent SDK, the Sandbox, and the self-hosted-vs-Managed-Agents decision. (Designed, not yet built.)