Markdown to PDF Converter

A sample document is loaded below. It is a live working example of every feature, so you can convert it as-is to see the output, or replace it with your own text. The Guide tab lists what each one does.

# FEATURE SHOWCASE - this line is Heading 1

This file exercises every markdown feature the converter supports, with each element
labelled so you can see what produced it. It doubles as the regression fixture: render it
after any change to the renderer and compare against the last known-good PDF.

Render it locally with:
`python scripts/render_local.py test/FEATURE_SHOWCASE.md C:/tmp/showcase`

## Headings - this line is Heading 2

### This line is Heading 3

#### This line is Heading 4

##### This line is Heading 5

###### This line is Heading 6

Heading 1 gets a box and a rule under it. Heading 5 and Heading 6 used to disappear
entirely; they render now.

## Text formatting

This sentence has **bold text**, *italic text*, `inline code` and ~~strikethrough~~ in it.
Bold and italic can be written with asterisks or with underscores: __also bold__ and
_also italic_.

You can combine them: ***bold and italic together***, and put `code` next to **bold**
without either one swallowing the other. Strikethrough survives inside **~~bold text~~**
too, and the tildes inside `a ~~code span~~` are left exactly as typed.

Links are live and clickable in the PDF: visit [tigzig.com](https://tigzig.com) or read
the [API catalog](https://tigzig.com/apis). An email link works too:
[write to Amar](mailto:amar@harolikar.com).

A link with no scheme, or an unsafe one, keeps its text but is not made clickable. That is
deliberate.

## Lists

### Unordered, with real nesting

- First item at the top level
- Second item, with **bold** inside it
  - Nested one level, written with two spaces
  - Another at the same level
    - Nested two levels deep
- Back to the top level

Each level gets its own marker and indent. Two-space indentation is what most people and
most AI tools write, and it nests correctly.

### Ordered, with real numbering

1. First step
2. Second step, with a [link](https://tigzig.com) in it
   1. Sub-step, numbered with letters
   2. Another sub-step
3. Third step

Top level numbers, the level below uses letters, the level below that uses roman numerals.

### A list glued to the line above

Summary of findings:
- Works even without a blank line before the list
- This is what people actually type
- The converter inserts the separation for you

### Task lists

- [x] Ticked items get a filled checkbox
- [x] Rendering fix, shipped
- [ ] Empty ones get an empty checkbox
- [ ] Images, still to come
- [ ] Task items can carry **bold** and [links](https://tigzig.com) like any other item

## Tables

Formatting survives inside table cells:

| Item | Status | Notes |
|------|--------|-------|
| Rendering fix | ✅ done | italics, links and `code` all render |
| Table alignment | ✅ done | `\|---:\|` right-aligns a column |
| Page break | ✅ done | see the ~~planned~~ page break section below |
| Task lists | ✅ done | real checkboxes, not literal brackets |
| Images | ❌ not yet | needs a security decision first |
| Emoji | ✅ done | mapped to symbols, never a black box |

A wider table still fits the page. The font and padding shrink as columns are added, so a
table can be cramped but it will never crash the conversion:

| Q1 | Q2 | Q3 | Q4 | FY | Growth | Margin | Share |
|----|----|----|----|----|--------|--------|-------|
| 120 | 135 | 128 | 150 | 533 | 12.4% | 18.2% | 4.1% |
| 98 | 104 | 119 | 122 | 443 | 8.9% | 15.7% | 3.4% |

## Code

Inline code looks like `df.groupby("state").sum()` inside a sentence.

A fenced block keeps its indentation:

```python
def convert(markdown_text: str) -> bytes:
    """Fenced code blocks render now - they used to vanish completely."""
    html = markdown_to_html(markdown_text)
    for line in html.splitlines():
        if line.strip():
            print(line)
    return build_pdf(html)
```

Other languages work the same way:

```sql
SELECT state, COUNT(*) AS filings
FROM   companies
WHERE  filed_on >= DATE '2026-01-01'
GROUP  BY state
ORDER  BY filings DESC;
```

## Blockquote

> A quote renders once, in its own style: italic, on a light background, indented on both
> sides. It used to lose its styling and come out looking like an ordinary paragraph.

## Emoji

Emoji have no glyphs in the standard PDF fonts, so they are mapped to real symbols drawn
from a bundled font. They never come out as black boxes.

- ✅ check, ❌ cross, ⚠️ warning, ⭐ star
- ⏳ pending, ⏸ paused, ⏹ stopped, ⚡ fast
- ➡️ right arrow, ⬅️ left arrow, ⬆️ up, ⬇️ down
- 👍 thumbs up, 👎 thumbs down, ❤️ heart, ✉️ mail
- 🔒 locked, 🔓 unlocked, 🔍 search, 📊 chart, 📈 up, 📉 down
- 🎯 target, ✏️ edit, ☎️ phone, ☀️ sun, ❄️ snow
- 🔴 🟢 🔵 coloured circles all become filled circles
- 🦆 🍕 🎉 anything decorative or unmapped becomes a small dot

An emoji only gets a symbol when the bundled font has a glyph that means the same thing.
A pin, a key or a light bulb have no faithful match, so they take the dot rather than a
substitute that would misread.

## Horizontal rule

The line below is a horizontal rule:

---

## Page break

Markdown has no way of saying "start a new page", which is the thing people most often want
from a PDF and cannot get. Put any one of these on a line by itself:

- `:::pagebreak:::`
- `\pagebreak` or `\newpage`
- ``
- `---pagebreak---`

They all do the same thing. Any of them inside a fenced code block is left alone, so the
text above stayed as text. The next line is a real one, so this sentence is the last thing
on its page.

:::pagebreak:::

## After the page break

This heading is at the top of a fresh page. Everything below continues normally.

## Special characters

Ampersands, angle brackets and quotes are escaped correctly: R&D spend, 5 < 6 > 4, and
"quoted text" with 'single quotes'.

Currency: ₹9,600 and ₹1,20,000 in rupees, plus £100, €50, $1.2M, ¥500 and ¢99. The rupee
sign has no glyph in the standard PDF fonts or in either bundled font, so it is written as
Rs. rather than left as a black box.

Maths: alpha ≤ 0.05, n ≥ 30, x ≠ y, ≈ 12%, √2, ∑ of returns, ∫ dx, ∂y/∂x, ∆ change and ∞.
These come from the Vera face that reportlab already ships, so they cost no extra font.

Already fine without help: 50% × 2 ÷ 4 ± 1, 5‰, ½ and ¾, µ, °C, © and ®.

A very long unbroken token such as
supercalifragilisticexpialidociousandthensomemoreletters wraps rather than running off the
page.

Greek renders correctly, upper and lower case: α β γ δ Δ σ Σ μ π λ ρ χ Ω. So do logic and
set symbols (∈ ⊂ ∪ ∩ ∀ ∃) and arrows in both weights (→ ← ↔ ⇒ ⇐ ⇔). ReportLab substitutes
its own Symbol face for these, so they need no extra font.

Non-Latin scripts (Devanagari, Arabic, CJK, Cyrillic) do NOT render and come out as filled
boxes. That one needs a bundled font per script, which is a separate decision.

## Not supported yet

These are known gaps, listed here so the fixture shows the honest current state:

- Images are skipped. A server that fetches image URLs supplied by a caller is a security
  question, so it needs a decision before it is built.
- Non-Latin scripts render as boxes, as noted above. Greek and maths are fine.

## Plain text files

A `.txt` upload is treated as plain text rather than markdown: its single line breaks are
kept, so a letter or a note does not collapse into one run-on paragraph. Markdown files
keep the standard behaviour, where a single newline is just a space.

The right-hand column shows the shape of the result, not the exact PDF.

What it does Type this You get

Text

Bold

**bold text**

bold text

Italic

*italic text*

italic text

Bold and italic

***both***

both

Strikethrough

~~removed~~

~~removed~~

Inline code

`df.head()`

df.head()

Link (clickable in the PDF)

[TigZig](https://tigzig.com)

TigZig

Headings

Six levels

# Heading 1
## Heading 2
### Heading 3
#### Heading 4
##### Heading 5
###### Heading 6

Heading 1

Heading 2

Heading 3

Heading 4

Heading 5

Heading 6

Lists

Bullets, nested with two spaces

- first
- second
  - nested
    - deeper

first
second
- nested
  - deeper

Numbered, nested becomes a. then i.

1. first
2. second
   1. nested

first
second
1. nested

Task list

- [x] done
- [ ] still to do

☑ done

☐ still to do

A list straight after a line of text, with no blank line

Summary:
- works anyway

Summary:

works anyway

Blocks

Table

| Fund | Return |
|------|--------|
| A    | 12.4%  |

Fund	Return
A	12.4%

Code block

```python
def run():
    return 1
```

def run():
    return 1

Quote

> a quoted line

a quoted line

Horizontal rule

---

Layout

Page break. Any of the five spellings, on its own line

:::pagebreak:::
\pagebreak
\newpage
<!-- pagebreak -->
---pagebreak---

everything after this starts on a new page

Plain .txt upload keeps its single line breaks

Dear Sir,
This stays on its own line.

Dear Sir,
This stays on its own line.

Symbols

Emoji become matching symbols, never a black box

✅ ❌ ⚠️ ⭐ 🎯 ➡️ 🔒 📈 ⏳

✔ ✗ ⚠ ★ ◎ ➡ 🔒 📈 ⌛

Rupee and other currency

₹9,600 · £100 · €50 · $1.2M

Rs.9,600 · £100 · €50 · $1.2M

Maths, logic and arrows

≤ ≥ ≠ ≈ √ ∑ ∫ ∂ ∞ ∈ ⊂ ∪ → ⇒ ↔

≤ ≥ ≠ ≈ √ ∑ ∫ ∂ ∞ ∈ ⊂ ∪ → ⇒ ↔

Greek letters, upper and lower case

α β γ δ Δ σ Σ μ π λ ρ χ Ω

α β γ δ Δ σ Σ μ π λ ρ χ Ω

Everything below is a known gap, listed so you find out here rather than inside a finished document. Nothing here fails the conversion.

Not supported	Example	What happens, and whether it is planned
Images	![caption](https://example.com/chart.png)	The image is skipped. Surrounding text is unaffected. Planned. Fetching image URLs supplied by a caller means our server makes outbound requests on request, so it needs a security decision first.
Hindi, Arabic, Chinese, Japanese, Korean, Cyrillic and other non-Latin scripts	नमस्ते · مرحبا · 你好 · Привет	Each character renders as a filled box. Latin text, Greek and maths in the same document are unaffected. On request. Each script needs its own font bundled, so we add one when there is a real need for it rather than guessing.
Raw HTML inside markdown	<div style="...">text</div>	The tag is ignored and its text is kept, so nothing is lost and nothing is styled. Not planned. Markdown formatting covers the same ground and this keeps the converter predictable.

Everything this page does is available as an API, and as an MCP server for AI agents. Two places to go, depending on who is reading.

For AI agents

A machine-readable description of this API lives in the agent brief, and it is listed in the API catalog (RFC 9727) alongside the other Tigzig APIs. The OpenAPI spec describes every endpoint, the Markdown features that are and are not supported, and the error contract.

Full documentation

Convert Markdown to a formatted PDF: send it as JSON or upload a file, and get the PDF back directly, or a hosted link if you are calling the MCP tool. Endpoints, examples, limits and the MCP setup are documented on tigzig.com/apis/md-to-pdf.