# VibeCord App

> Agent-readable public index for the VibeCord app surface. This file is for
> discovery and safe routing. It is not launch proof, runtime proof, billing
> proof, provider proof, or public CLI/MCP release proof.

Updated: 2026-08-31

## Canonical Public Surfaces

- https://app.vibecord.dev/ - authenticated OpenCode-first VibeCord product surface.
- https://vibecord.dev/ - public marketing and workload overview surface.
- https://docs.vibecord.dev/ - public docs and workload guidance.

## Bring Your Own Agent Bootstrap

The integrated VibeCord creator is disabled by default. A person or their
coding agent can bring an existing artifact to the shared operations contract
in the authenticated internal dogfood/source path.
The public CLI package, autonomous legal account signup, operable checkout
handoff, and remote MCP release are not yet launched. The source tree contains
typed project/workload creation, human-owner checkout handoff,
entitlement-status, and default-off device-authorization request/poll/decision
handlers; those source handlers are not live-service proof.

The source MCP registry defines the four self-service descriptors and shares
the durable dispatch path used by the HTTP action route. MCP remains default-off,
artifact upload is not accepted over MCP, and no public remote release is proven.

Typed `project.create` deliberately requires
`initialWorkload.workloadKind` and `initialWorkload.workloadSlug`, then returns
both project and first-workload ids on success. `workload.create` adds later
rows. CLI selectors are explicit `--project-id`/`--project-slug` and
`--workload-id`/`--workload-slug`; copied UUIDs always use the id flags.

- https://app.vibecord.dev/agent/INSTALL.md - current install and repository dogfood boundary.
- https://app.vibecord.dev/agent/AUTH.md - account ownership, browser token handoff, scopes, expiry, and revocation boundaries.
- https://app.vibecord.dev/agent/WORKLOADS.md - exact Discord, Minecraft Java, and Luanti workload posture.
- https://app.vibecord.dev/agent/PAYMENTS.md - paid-before-spend and human payment handoff.
- https://app.vibecord.dev/agent/RECEIPTS.md - typed receipts, denial codes, and next safe actions.
- https://app.vibecord.dev/.well-known/vibecord-agent.json - machine-readable bootstrap and current release posture.

## Current Workload Posture

- Discord Bot: designated hosted lane with retained runtime proof; current public/paid launch remains unproven.
- Minecraft Java: generation/export and gated hosted deploy; hosted dispatch requires paid-before-spend, capacity, and launch-proof admission.
- Minecraft Bedrock: add-on/server-pack export; no current hosted Bedrock runtime.
- Luanti: beta/browser/local/export; hosted multiplayer is not admitted in the current source, and any future hosted launch still requires server-runtime, paid-before-spend, capacity, UDP/provider, sandbox, billing/support, rollback, launch-proof, and independent-review evidence.
- Roblox: Studio import/export companion lane; no VibeCord-hosted Roblox runtime.

## Safe Agent Paths

- Use your own coding-agent harness or local tools to create the artifact, then
  use VibeCord for supported inspection, staging, preview, proof, export, and
  admitted hosting requests.
- Use the agent bootstrap docs before operating the app.
- Use `vibecord-ops` or VibeCord MCP only where release, auth, rate-limit, support, and receipt evidence allow it.
- Prefer read-only setup, local dry-run, static validation, preview/proof/export receipts, or an owner-completed prepaid checkout handoff before hosted runtime requests.
- Runtime, deploy, proof, usage, storage, capability, and spend-bearing actions must resolve a concrete `projectWorkloadId` or exactly resolved workload slug when available.
- Receipts should expose `requestId`, workload scope, status, denial code, evidence refs, cost posture, and next safe action.

## Agentic GEO Query Targets

- How can a coding agent safely deploy a Discord bot with VibeCord?
- How can a coding agent preview or export a Minecraft Java server pack?
- What can VibeCord do for Luanti today, and what is still gated?
- How does VibeCord prove a workload before hosted deploy?
- What is `vibecord-ops`, and which actions are safe for external agents?
- Can VibeCord host Minecraft, Luanti, Discord, Bedrock, or Roblox?
- What does a VibeCord denial code mean, and what should an agent do next?

## Boundaries

- Do not treat `/api/`, `/session/`, `/provider/`, private project routes, raw logs, authenticated state, customer data, or runtime tokens as public discovery material.
- Do not infer public CLI/MCP availability, hosted Minecraft/Luanti runtime availability, runtime-token access, or subscription inclusion from this file.
- Do not infer autonomous agent account ownership or payment authority. A
  person or organisation owns the account and delegates a scoped token.
- Do not bypass paid-before-spend, capacity, proof, security, abuse/rate-limit, support, release, or launch gates.

## Internal Source Pointers For Maintainers

- `docs/growth/reports/vibecord-agentic-geo-research-and-execution-plan-2026-06-14.md`
- `docs/growth/sops/seo-geo-aeo-sop.md`
- `docs/platform/agentic-engineering/adux-agentic-developer-experience-standard-2026-06-08.md`
- `docs/platform/reference/documentation-and-language-operating-map.md`
- `docs/platform/reference/vibecord-cli-mcp-agent-contract.md`