# Inkroost

> Inkroost is a premium manga publishing platform where independent creators publish original manga series, and readers discover, read, and support the artists behind them.

## What is Inkroost?

Inkroost lets creators produce AI-assisted manga series — including story structure, character sheets, AI-generated cover art, and chapter publishing — all within a single platform. Readers can browse thousands of series, read free chapters, and unlock premium content with credits.

## AI Agent Integration (MCP)

Inkroost exposes a Model Context Protocol (MCP) server that lets AI agents create and manage manga content on behalf of a creator. Agents can plan a series, build character sheets, generate AI art, publish chapters, and track credits — all programmatically.

The server implements the MCP 2025-03-26 specification over HTTPS (Streamable HTTP transport).

### Endpoint

A single base URL serves every tool and resource:

```
https://inkroost.com/mcp
```

### Connecting

Two paths work today. Pick the one your client supports.

**Option 1 — Custom Connector (recommended for Claude.ai)**

Claude.ai (Pro / Team / Enterprise) signs in via OAuth and gets full access:

1. In Claude, open the left sidebar → **Customize** → **Connectors**
2. Click the **+** button → **Add custom connector**
3. Paste `https://inkroost.com/mcp` and click **Add**
4. Sign in to Inkroost when prompted, then click **Allow**

The connector then shows every tool in the catalog below.

**Option 2 — Bearer token (Claude Code, Cursor, Windsurf, OpenAI Agents, scripts)**

1. Log in at https://inkroost.com/login
2. Visit https://inkroost.com/settings/api-tokens
3. Create a token with the abilities your agent needs
4. Send `Authorization: Bearer <token>` with every request

Example for Claude Code:

```
claude mcp add --transport http inkroost https://inkroost.com/mcp \
  --header "Authorization: Bearer YOUR_TOKEN_HERE"
```

### Token abilities (manual tokens only)

| Ability | What it unlocks |
|---------|-----------------|
| `story:read` | Read series, story state, characters, consistency context |
| `story:write` | Create / update series, characters, chapters, style profiles |
| `art:generate` | Generate AI art (charges credits per generation) |
| `publish:write` | Publish chapters and submit for review |
| `credits:read` | Check balance, estimate costs, list packages |
| `credits:purchase` | Initiate credit purchases (returns a checkout URL the user opens) |

OAuth-issued tokens (Option 1) grant the umbrella `mcp:use` scope, which covers every tool.

### Tools

26 tools, grouped by area:

**Story**
- `create-series-tool` — Create a new manga series
- `set-series-style-profile-tool` — Set the series-wide default visual style profile. Call this immediately after `create-series-tool` and before any page generation — the profile is auto-injected into every generation prompt.
- `get-story-state-tool` — Read the current state of a series (chapters, characters, metadata)
- `get-consistency-context-tool` — Inspect the style profile and per-character completeness scores that will be injected into generation prompts
- `create-chapter-tool` — Create a new chapter shell within a series
- `update-page-tool` — Update an existing chapter's title, description, status, or position

**Characters**
- `create-character-tool` — Create a character sheet
- `update-character-tool` — Update an existing character sheet
- `set-character-style-profile-tool` — Fix this character's visual identity (hair, eyes, marks, outfit, build) so they render consistently across every page they appear in
- `get-character-sheet-tool` — Retrieve a character sheet

**Art**
- `generate-cover-tool` — Generate an AI cover image
- `generate-banner-tool` — Generate an AI banner image (16:9)
- `generate-pages-tool` — Generate manga page variants for a chapter. Series and referenced character style profiles are auto-injected for consistency. By default no in-image text is rendered.
- `generate-pages-batch-tool` — Queue multiple page-generation jobs in one call (up to 20 prompts per batch)
- `get-art-status-tool` — Poll a generation job and fetch result URLs once ready
- `get-daily-limit-status-tool` — Check today's AI generation usage and remaining capacity
- `approve-pages-tool` — Approve generated variants and create Page records (per-variant selection or bulk discard)
- `batch-approve-pages-tool` — Approve multiple generation jobs at once (up to 50 job IDs)
- `annotate-page-tool` — Composite readable dialog, captions, or narration onto an approved page

**Publish**
- `publish-chapter-tool` — Publish a chapter (returns a warning if the chapter has fewer than 5 pages — manga chapters typically run 8-24 sequential pages)
- `submit-for-review-tool` — Submit a chapter for moderation review
- `check-approval-status-tool` — Check the approval status of a chapter

**Credits**
- `check-balance-tool` — Current credit balance
- `estimate-generation-cost-tool` — Estimate credits required for AI generation
- `list-credit-packages-tool` — List available credit packages (ids + prices)
- `purchase-credits-tool` — Initiate a checkout session for a package; returns a URL the user opens

**Daily generation limit:** 500 AI images per user per day, shared across every generation tool. Counter resets at midnight UTC. Call `get-daily-limit-status-tool` before queuing large batches.

## Recommended Agent Workflow

1. `create-series-tool` — make the series.
2. `set-series-style-profile-tool` — set the default visual style (subject + photography + background + constraints). This is the single biggest lever for consistency across pages.
3. For each named character: `create-character-tool` → `set-character-style-profile-tool` with detailed appearance (hair length + colour + parting, eye colour + shape, distinguishing marks, default outfit, build). Vague profiles produce inconsistent art.
4. `create-chapter-tool` — make a chapter shell.
5. `generate-pages-tool` per beat (or `generate-pages-batch-tool` for multiple beats in one call) — manga chapters are multiple sequential pages telling one sub-story, typically 8-24 pages. Pass `character_ids` so the right style profiles get used.
6. `get-art-status-tool` to poll; fetch and inspect each result URL before approving.
7. `approve-pages-tool` / `batch-approve-pages-tool` to commit the best variants as Page records.
8. `publish-chapter-tool` — only after every page is approved.

Before generating, call `get-consistency-context-tool` to see the current style profile and per-character completeness scores. Self-correct any low scores first.

## Character Definitions — Be Hyper-Specific

AI image models have no persistent memory between generations. Set rich character style profiles **once** via `set-character-style-profile-tool` so every subsequent generation that references the character gets the same anchors:

- **Hair:** colour, length, style (e.g. "jet-black straight hair, blunt jaw-length cut, fringe swept left")
- **Eyes:** colour, shape, distinctive features (e.g. "pale ice-blue almond-shaped eyes, thin dark limbal ring")
- **Build:** approximate height, body type, muscle definition
- **Skin tone:** descriptive (e.g. "warm golden-brown")
- **Signature outfit:** specific garments and colours
- **Unique markings:** scars, tattoos, accessories that appear consistently

Vague descriptions (e.g. "a girl with dark hair") produce inconsistent characters across pages.

## Review Before Approving

When `get-art-status-tool` returns `result_urls`, fetch and analyse each image before approving:

1. Are all characters recognisable and consistent with their style profiles?
2. Does the scene match the intended composition, action, and mood?
3. If `allow_text=true` was used, is dialog readable and correctly attributed?

If the image doesn't meet expectations, call `approve-pages-tool` with `discard: true` and regenerate with a refined prompt. Iterative refinement produces higher-quality pages — treat generation as a loop, not a one-shot operation.

## Using `allow_text`

Only enable `allow_text: true` when you want speech bubbles handled by the AI model. When you do:

- Provide an exact dialog script in reading order
- Name every speaker by character name
- Use the format `Panel N — Speaker: "dialog text"` — speaker prefixes are stripped automatically before the bubble is rendered

For precise control over dialog, leave `allow_text: false` and apply `annotate-page-tool` afterwards instead.

See the full agent guide at https://inkroost.com/llms-full.txt for detailed tool schemas, inputs, outputs, and workflow examples.

## Sitemap

- Homepage: https://inkroost.com/
- Browse Series: https://inkroost.com/browse
- Pricing: https://inkroost.com/pricing
- Creator Dashboard: https://inkroost.com/creator (authenticated)
- AI Agent Access: https://inkroost.com/settings/api-tokens (authenticated)
- Sitemap: https://inkroost.com/sitemap.xml
- Full agent docs: https://inkroost.com/llms-full.txt