# MCP server

Use SciFig from Claude, ChatGPT, Claude Code, Codex and Cursor through the remote MCP server.

Source: https://scifig.ai/docs/api/mcp

## MCP server

Server URL: `https://mcp.scifig.ai/mcp` (Streamable HTTP). Clients that support OAuth connect by signing in to SciFig and approving access; clients that send headers can use an API key as `Authorization: Bearer sk_live_…`.

| Tool | What it does |
| --- | --- |
| generate_figure | Create a scientific illustration from a prompt |
| edit_figure | Edit an image: instruction (optionally a masked region), white background, aspect ratio, recolor or text replacement |
| detect_text | Find the text in an image before replacing it |
| upscale_figure | Upscale to 2K, 4K or 8K |
| convert_figure | Convert to SVG, PNG, PDF, TIFF or PPTX |
| generate_chart | Plot a chart from CSV or TSV text or JSON rows (up to 1 MB) |
| edit_chart | Change the latest version of a chart |
| generate_flowchart | Draw a flowchart as SVG, optionally from up to 8 attachments |
| edit_flowchart | Change the latest version of a flowchart |
| get_job | Status, credits and download links of a job |
| list_figures | Recent jobs |
| get_credit_balance | Credit balance, plan and limits |

Creation tools wait up to 25 seconds. If a job is still running, call `get_job` later; calling the creation tool again creates and charges a new job. Results list each download link and, for charts and flowcharts, the link that opens the result in the SciFig editor.

## Connect your client

Claude (claude.ai and Claude Desktop): Settings → Connectors → Add custom connector, enter the server URL, then sign in to SciFig and allow access.

ChatGPT: Settings → Apps → Advanced → Developer mode, create an app with the server URL and OAuth authentication.

Claude Code:

```bash
claude mcp add --transport http scifig https://mcp.scifig.ai/mcp
```

Codex (`~/.codex/config.toml`):

```toml
[mcp_servers.scifig]
url = "https://mcp.scifig.ai/mcp"
```

Cursor (`~/.cursor/mcp.json`), with an API key from Settings → Developer:

```json
{
  "mcpServers": {
    "scifig": {
      "url": "https://mcp.scifig.ai/mcp",
      "headers": { "Authorization": "Bearer sk_live_…" }
    }
  }
}
```

Clients without OAuth can add the header `Authorization: Bearer sk_live_…` with an API key from Settings → Developer.

## Tool parameters

| Tool | Parameter | Required | Rules |
| --- | --- | --- | --- |
| generate_figure | prompt | yes | 3–20,000 characters |
|  | aspect_ratio | no | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9` |
|  | quality | no | `standard`, `advanced` or `premium` |
|  | style | no | One of `flat`, `2-5d`, `line-art`, `hand-drawn`, `scientific`, `line`, `3d`, `editorial`, `sketch`, `watercolor`, `auto` |
|  | palette | no | A palette name of 1–64 characters, or 1–4 `#RRGGBB` colours |
|  | reference_images | no | Up to 3 images: file_id, https URL or data URL; PNG, JPEG or WebP up to 20 MB (data URL about 3 MB, see below) |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| edit_figure | image | yes | file_id, https URL or data URL; PNG, JPEG or WebP up to 20 MB (data URL about 3 MB, see below) |
|  | operation | no | `instruct` (default), `white_background`, `aspect_ratio`, `recolor` or `replace_text` |
|  | instruction | for `instruct` | 1–2,000 characters |
|  | mask | no | Only for `instruct`: a PNG up to 8 MB, white where the image may change |
|  | aspect_ratio | for `aspect_ratio` | `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9` |
|  | palette | no | Only for `recolor`: 1–8 `#RRGGBB` colours |
|  | text_detection | for `replace_text` | The id returned by `detect_text` |
|  | replacements | for `replace_text` | 1–256 items `[{"index", "text"}]`: index 0–255 from `detect_text`, each index at most once; text 1–500 characters |
|  | quality | no | Ignored by `replace_text` |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| detect_text | image | yes | file_id, https URL or data URL; PNG, JPEG or WebP up to 20 MB (data URL about 3 MB, see below) |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| upscale_figure | image | yes | file_id, https URL or data URL; PNG, JPEG or WebP up to 20 MB (data URL about 3 MB, see below) |
|  | scale | yes | `2k`, `4k` or `8k` |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| convert_figure | image | yes | file_id, https URL or data URL; PNG, JPEG or WebP up to 20 MB (data URL about 3 MB, see below); or a flowchart’s SVG output by file_id |
|  | format | yes | `svg`, `png`, `pdf`, `tiff` or `pptx` |
|  | vector_mode | no | Only with format `svg`: `editable_text` or `full_vector` |
|  | dpi | no | Not for `svg`: 300, 600 or 1000 |
|  | width | no | Not for `svg`; only for flowchart SVG sources: 256–4096 pixels |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| generate_chart | data | data or rows | CSV or TSV text with a header row, up to 1 MB (1,000,000 characters) |
|  | rows | data or rows | Instead of `data`: 1–100,000 objects, column name to value; sent as CSV, up to 1 MB |
|  | data_format | no | `csv` or `tsv`; detected from the first line when omitted |
|  | request | yes | What to plot, 1–16,000 characters |
|  | chart_types | no | Up to 46 chart types, no duplicates (see [charts](/docs/api/charts)) |
|  | journal_style | no | `default`, `nature`, `science`, `cell`, `plos` or `acs` |
|  | figure_width | no | `single`, `onehalf` or `double` |
|  | color_palette | no | A palette name (see [charts](/docs/api/charts)) |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| edit_chart | job_id | yes | The latest chart job |
|  | instruction | yes | 1–16,000 characters |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| generate_flowchart | prompt | yes | 1–8,000 characters |
|  | aspect_ratio | no | `auto`, `16:9`, `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `21:9` |
|  | attachments | no | Up to 8: file_id, https URL or data URL; PDF, DOC, DOCX, Markdown, text, SVG, PNG, JPEG or WebP, up to 32 MB each |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| edit_flowchart | job_id | yes | The latest flowchart job |
|  | instruction | yes | 1–8,000 characters |
|  | idempotency_key | no | Optional; reuse only to retry the same request after a timeout or network error |
| get_job | job_id | yes | A job id |
| list_figures | limit | no | 1–20 |
|  | cursor | no | `next_cursor` of the previous page |
| get_credit_balance | — | — | No parameters |

`generate_chart` takes text data only, up to 1 MB. For Excel workbooks (and their `sheet_name`) or larger files, use [`POST /charts/generations`](/docs/api/charts).

## How the tools behave

- The server is `https://mcp.scifig.ai/mcp` (Streamable HTTP). It calls this API on your behalf, so prices, credits, limits and errors are the same.
- Image and attachment arguments accept a `file_id`, an https URL or a data URL; URLs and data URLs are uploaded first with `POST /files`, so they count against the upload limits. SciFig’s own URLs are refused (pass the `file_id`), and redirects are followed only to other https addresses.
- A request to the MCP server is limited to 4 MB. Base64 makes a file a third larger, so a data URL can carry about 3 MB; pass larger files as an https URL.
- The server checks your credential only on `initialize` and `tools/call`. An expired or revoked token returns `401` so the client can refresh or sign in again; a banned account, or one being deleted, returns `403`.
- Finished image results are shown inline when they are at most 1 MB; every result includes download links, and chart and flowchart results add the link that opens them in the SciFig editor.
- Every creation uses a new Idempotency-Key unless the assistant passes `idempotency_key`, so asking again creates and charges a new job. Use `get_job` to follow up.
- An error result echoes `idempotency_key` and, when a job exists, `job_id`. After a timeout or a server error, retry with the same key or call `get_job`; a refusal such as `402 INSUFFICIENT_CREDITS` stays stored under its key, so send it again with a new key.