Use `/agents` for product evaluation
When the question is what Move Big Rocks is, what it replaces, or whether the platform is a fit, start at `/agents` and `/docs/cli` instead.
The SaaS-replacement platform
Shared system for teams and agents. Official CLI + GraphQL surface.
Extension developers
Give Claude Code, Codex, or any capable agent the extension SDK and describe what you want. It scaffolds, implements, tests, and deploys a working extension against the same runtime model as the first-party set. No team required. No months of development. This is the canonical public handoff for humans and agents building a custom Move Big Rocks extension. If the task is product evaluation, start at `/agents` instead. If there is no running instance yet, stop and use `/docs/self-host` first.
Table of contents
Jump directly to the part you need.
When to use this route
The path is easier for humans and much easier for agents when the first routing decision is explicit. Do not mix evaluation, deployment, and extension authoring into one reading order.
When the question is what Move Big Rocks is, what it replaces, or whether the platform is a fit, start at `/agents` and `/docs/cli` instead.
If there is no running instance or preview workspace, do not start from the SDK. Create the owned runtime first.
If the workflow can already live on workspaces, teams, forms, queues, cases, knowledge, and automation, use the shared base first. Many approvals and requisitions are configuration work, not extension authoring.
When the job is to build, upgrade, review, or explain a private extension, start here and then move deeper into the build and runtime references.
Agent reading order
The goal is a deterministic route an agent can follow without having to guess which page matters next.
Confirm prerequisites, the supported self-built slice, and whether the task is really extension work.
Use the detailed build guide for the full lifecycle, review gates, and command loop.
Use `README.md` and `START_HERE.md` in the public SDK as the concrete scaffold and local workflow.
Proof
Branch into service-backed runtime design only when the extension truly needs handlers, jobs, event consumers, or an owned schema.
Proof
Use the extension model reference when runtime class, endpoint class, storage class, or trust boundaries are in question.
Supported self-built slice
The SDK and site now say this plainly because agents were otherwise too likely to infer an overpowered starting point.
Proof bundle
Default proof loop
This is the shortest supported path from a private extension repo to a preview-workspace extension with an inspectable surface.
Proof bundle
Use this exact loop before you tell anyone the extension is ready.
$ git clone https://github.com/MoveBigRocks/extension-sdk acme-my-extension$ cd acme-my-extension$ mbr extensions lint . --json$ mbr auth login --url https://app.yourdomain.com$ mbr extensions verify . --workspace ws_preview --json$ mbr extensions nav --instance --json$ mbr extensions widgets --instance --json$ mbr extensions skills list --id EXTENSION_ID --json
Service-backed branch
Service-backed runtime is a legitimate next step, but it should be a conscious upgrade rather than the default starting point.
Proof bundle
Branded pages, admin UI, bundled skills, workflow seeds, and declarative product slices usually do not need a separate backend process.
Choose `service_backed` when the extension needs custom server-side handlers, public ingest, or compatibility-sensitive endpoints.
Choose `service_backed` when the extension needs scheduled jobs, domain event consumers, or richer background work.
Choose `service_backed` when the extension needs an `ext_*` PostgreSQL schema, its own migrations, and a bounded data model.
Common failure modes
The extension author route should now be explicit enough that these failure modes are unnecessary.
References
Each link goes to the next authoritative page, reference, or support surface.
Move Big Rocks
Start from /agents, use
/docs/cli as the official product
tour, inspect /resources for
source and proof, and review /security
before making deployment or data-handling decisions.