# SpriteShip

> SpriteShip turns text prompts into production-ready 2D game art — characters with
> animated spritesheets, tilesets, terrain, seamless textures, backgrounds, and full
> levels — and exports them to Phaser, Godot, Unity, GameMaker, and Tiled. It is built
> to be driven by an AI coding agent: everything the web app can do, an agent can do
> over a hosted MCP server or a REST API, under spend controls the agent cannot raise.

## Connect an agent

**MCP (recommended).** Streamable HTTP endpoint:

    https://spriteship.com/mcp

Authenticate with an API key as a Bearer token. Example client config:

    {
      "mcpServers": {
        "spriteship": {
          "type": "http",
          "url": "https://spriteship.com/mcp",
          "headers": { "Authorization": "Bearer ss_live_..." }
        }
      }
    }

Claude Code one-liner:

    claude mcp add --transport http spriteship https://spriteship.com/mcp \
      --header "Authorization: Bearer ss_live_..."

**REST.** The same surface as plain HTTP, for agents without MCP support:

    curl -H "Authorization: Bearer ss_live_..." https://spriteship.com/api/v1/me

Create a key at https://spriteship.com/account/api-keys. The raw key is shown once.

## Spending is gated, and the gate is not advisory

Generations cost credits — real money. Four things are true of every key, and none
of them can be changed by the key itself; only the account owner can, from the
dashboard. An agent that has been prompt-injected cannot raise its own budget.

- **Scopes are fixed at creation.** `assets:read` and `account:read` are the
  defaults. `assets:write` adds free, no-AI operations (frame curation,
  duplicating an animation, promoting a pose to a still, cancelling a job).
  `credits:spend` is a separate, explicit opt-in and is the only scope that can
  cost money.
- **Every paid operation can be priced before it runs.** Over MCP, paid tools
  return a cost quote by default and execute nothing until the agent explicitly
  opts in. Over REST, add `?dryRun=true` to any paid POST. Either way you get
  `estimatedCredits`, the current balance, and the key's remaining daily headroom
  without spending anything.
- **Every key has a daily spend cap**, defaulting to 5,000 credits per rolling
  24 hours, plus a paid-attempt ceiling. Accounting is gross: refunds restore the
  balance but never restore the cap's headroom, so a key whose generations keep
  failing exhausts its own budget instead of looping forever.
- **Paid writes require an idempotency key**, so a retry after a network failure
  cannot double-charge.

Reads are free on every plan, and structurally so: the read surface never touches
the credit ledger.

## What an agent can do

- **Read the library** — projects, characters (the full Phaser-ready JSON with
  animations, atlases, mounts, and poses), assets of all six types, maps with their
  saved level data, custom-field definitions, credit balance and per-key ledger.
- **Plan a safe re-sync** — compare a local lock file in one free call, receive
  field-level summaries for represented animation/collision/runtime/map fields,
  a coarse signal for other content, and review unavailable entities before any
  local file is removed.
- **Export** — engine bundles for Phaser, Godot, Unity, GameMaker, and Tiled maps.
  Every zip carries a self-describing bundle (a manifest, a JSON Schema, a README,
  and a SKILL.md) so an agent can wire the output into a game project unaided.
- **Curate for free** — pick which frames an animation keeps, duplicate its current
  saved state exactly, promote a pose into a 1-frame idle, cancel a job.
  None of these touch the ledger.
- **Verify generated animation delivery** — a completed animation job requires a
  character quality read; machine-readable gates lead the agent through free
  frame inspection/repair before it can call the animation ready.
- **Generate, with consent** — new character poses, animate any asset, revise an
  asset, re-roll a single motion clip, make a texture tile seamlessly.

## Documentation

- [API Overview](https://spriteship.com/docs/api-overview): auth model, scopes, the read/write split, and how paid calls are gated
- [API Reference](https://spriteship.com/docs/api-reference): every /api/v1 endpoint with parameters and response shapes
- [Errors & Limits](https://spriteship.com/docs/api-errors-and-limits): the machine-readable error `code` table, rate-limit tiers, and durable ceilings
- [Use with Claude Code](https://spriteship.com/docs/agent-skill): the distributable agent skill and its slash commands
- [Account & API Keys](https://spriteship.com/docs/account-and-api-keys): creating an ss_live_ key, choosing scopes, setting a per-key daily spend cap
- [Credits & Billing](https://spriteship.com/docs/credits-and-billing): what a generation costs and how refunds work
- [Exports](https://spriteship.com/docs/exports): the self-describing export bundle every zip ships
- [Getting Started](https://spriteship.com/docs/getting-started): what SpriteShip produces, end to end
- [Core Concepts](https://spriteship.com/docs/core-concepts): projects, characters, animations, assets, maps
- [Characters](https://spriteship.com/docs/characters): character generation, poses, and rotation
- [Animations](https://spriteship.com/docs/animations): how animations are generated, curated, and normalized
- [Assets](https://spriteship.com/docs/assets): the six asset types and what each produces
- [Animating Assets](https://spriteship.com/docs/animating-assets): animation as a capability of every asset type
- [Level Editor](https://spriteship.com/docs/level-editor): placing assets into maps
- [Custom Fields](https://spriteship.com/docs/custom-fields): typed per-entity and per-placement data baked into exports
- [Prompting Guide](https://spriteship.com/docs/prompting-guide): what wording actually changes the output

## Also

- [Marketplace](https://spriteship.com/marketplace): free, royalty-free asset packs; every pack is a real export bundle
- [Pricing](https://spriteship.com/pricing): credit packs and subscriptions
- [Features](https://spriteship.com/features): the full capability list