# ganjoor-mcp-server [Health: Active]

**Category:** 🎨 Art & Culture  
**Repository:** https://github.com/ganjoor/ganjoor-mcp-server  
**GitHub Stars:** 7  
**npm Downloads (last month):** 299  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/ganjoor-mcp-server

## Description
An MCP server for Ganjoor the largest public archive of Persian classical poetry — exposing the read-only public API  to AI agents.

## Tools
Capabilities this server exposes over MCP:

- **ganjoor_list_poets** — List all poets published on Ganjoor.net, including their biographies.

Returns one row per poet with: id, name, nickname, URL slug, root category id, birth/death years
(Hijri), and birth/death places. Use this first when you need to translate a poet's name to a
numeric id for other Ganjoor tools.

Args:
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  A paginated poet list. For JSON: { total, count, page, page_size, has_more, next_page, poets: [{ id, name, nickname, full_url, root_cat_id, birth_year_hijri, death_year_hijri, birth_place, death_place }] }

Examples:
  - "Who is on Ganjoor?" -> page_size=50
  - "List the poets born in the 10th century Hijri" -> fetch pages and filter birth_year_hijri

Notes:
  - Only public, read-only data. No authentication required.
  - The list is long (~170 poets); paginate rather than requesting page_size=100 repeatedly.
- **ganjoor_get_poet** — Get a single poet's full profile (including biography) by numeric id or URL slug.

Use this to turn a slug like "/hafez" into the numeric poet id, or to read a poet's biography
and catalogue of works.

Args:
  - id (number): Numeric poet id (optional if `url` is given). Hafez = 2, Saadi = 7, Rumi = 5.
  - url (string): Poet URL slug such as 'hafez' or '/hafez' (optional if `id` is given)
  - include_cats (boolean): Include the poet's top-level book/section list (default: true)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  Poet profile with id, name, nickname, biography, birth/death (Hijri + places), plus for JSON:
  { id, name, nickname, biography, birth_year_hijri, death_year_hijri, birth_place, death_place, full_url, cats: [{ id, title, full_url, cat_type }] }

Examples:
  - "Tell me about Hafez" -> url='hafez'
  - "What are Saadi's works?" -> id=7, include_cats=true

Notes:
  - Supply exactly one of `id` or `url`. If both are given, `id` wins.
- **ganjoor_get_poets_by_century** — Get every Ganjoor poet grouped into historical half-centuries.

The first group (id 0) holds the "pinned"/featured poets shown on the Ganjoor home page. Each later
group carries a name, an approximate Gregorian year range, and the poets active in that period.

Args:
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { centuries: [{ id, name, start_year, end_year, poet_count, poets: [{ id, name, full_url }] }] }

Examples:
  - "Which poets are from the 14th century?" -> inspect the century entries and their poets

Notes:
  - The whole payload is returned at once; it is a few hundred KB at most.
  - Grouping is editorial, so a poet appears exactly once.
- **ganjoor_list_books** — List Ganjoor books (named collections such as دیوان حافظ or غزلیات سعدی), alphabetically.

Every returned entry is a category of type "Book". Use the resulting category id with
ganjoor_get_category to browse the poems inside, or with ganjoor_get_category_poems to list them.

Args:
  - name (string): Optional filter — only books whose name contains this substring (case-insensitive)
  - poet_id (number): Optional filter — only books belonging to this poet id
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, books: [{ id, name, poet_id, poet_name, full_url }] }

Examples:
  - "What books does Hafez have?" -> poet_id=2
  - "Find books matching 'ghazal'" -> name='ghazal'
  - "Show me Molavi's collections" -> poet_id=5

Notes:
  - Filtering happens server-side, so it is much cheaper than fetching everything and filtering locally.
  - Ganjoor's book catalog is a curated subset: some major poets (including Hafez) have no entries.
    For them use ganjoor_get_poet with include_cats=true, which lists a poet's sections directly.
- **ganjoor_get_category** — Get a Ganjoor category (a book, section, or sub-collection) by numeric id.

A category is the container for poems inside a poet's work — for example Hafez's ghazals
(category id 10) or Rumi's Masnavi. The response includes the breadcrumb path, the direct
children (sub-sections), and optionally the poems it contains.

Args:
  - cat_id (number): Category id. Hafez's root category is 9.
  - include_poems (boolean): Include the poems inside this category (default: true)
  - include_children (boolean): Include the direct child categories/sections (default: true)
  - poem_limit (number): Maximum poems to return when include_poems is true, 1-200 (default: 50)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  Category metadata: id, title, breadcrumb, description, table-of-contents style, child categories,
  and (optionally) contained poems with their ids, titles and URLs. Set include_poems=false to skip
  the poem list; use ganjoor_list_category_poems when you only want the poems and their total count.

Examples:
  - "What is inside Hafez's ghazals?" -> cat_id=10, include_children=false
  - "List Hafez's sections" -> cat_id=9, include_poems=false

Notes:
  - A category that directly contains poems returns those poems; a pure section returns children
    instead. Follow `children` to descend.
- **ganjoor_list_category_poems** — List the poems contained in a Ganjoor category, optionally filtered by the author's name.

Args:
  - cat_id (number): Category id to list poems from
  - title_contains (string): Optional filter — only poems whose title contains this text
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Poems per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { total, count, page, page_size, has_more, next_page, category: { id, title, full_url }, poems: [{ id, title, url_slug, full_url, excerpt }] }

Examples:
  - "Which poems are in category 10?" -> cat_id=10
  - "Find the 'شمارهٔ ۱' poems in a category" -> title_contains='شمارهٔ ۱'

Notes:
  - Title filtering is done client-side because the upstream category endpoint is unpaged.
  - For large collections, combine this with page_size to walk the whole diwan.
- **ganjoor_get_category_geotags** — Get the geographic and historical locations tagged to poems within a Ganjoor category and its descendants.

Each tag pairs a place with the Hijri/Gregorian date a poem commemorates, which is how Ganjoor
records historical events (a battle, a siege) in poetry.

Args:
  - cat_id (number): Category id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { total, tags: [{ id, latitude, longitude, lunar_date, gregorian_date, child_cat }] }

Examples:
  - "Where are the events in Hafez's poetry set?" -> cat_id=9

Notes:
  - Tag lists can be large (tens of entries for a whole diwan).
  - Coordinates are decimal degrees, WGS84.
- **ganjoor_get_category_people_graph** — Get the network of historical people relevant to a Ganjoor category or work.

Returns every person tagged in a poem under this category's subtree, plus their kinship and
affiliation ties one hop out — the data behind Ganjoor's "characters" (شخصیت‌ها) tab.

Args:
  - cat_id (number): Category id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { node_count, edge_count, nodes: [{ id, name, importance }], edges: [{ from_person_id, to_person_id, relation }] }

Examples:
  - "Which historical figures appear in the Shahnameh?" -> cat_id=<Shahnameh category id>

Notes:
  - Empty graphs are normal for categories with no person tagging — that is not an error.
- **ganjoor_search_poems** — Full-text search across all poems on Ganjoor, optionally scoped to one poet or category.

The search matches Persian text inside poem bodies and titles. Use this when you know what the
poem says but not what it is called; use ganjoor_list_category_poems when you know the book.

Args:
  - term (string): Search text. Persian phrases work best; matches are substring based.
  - poet_id (number): Restrict to one poet (default: 0 = all poets)
  - cat_id (number): Restrict to one category (default: 0 = all categories)
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Results per page, 1-100 (default: 20)
  - include_text (boolean): Include each poem's full text in the response (default: false)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, poems: [{ id, title, full_title, full_url, plain_text? }] }

Examples:
  - "Find poems mentioning 'دل'" -> term='دل', poet_id=2
  - "Search Hafez for 'صبر'" -> term='صبر', poet_id=2
  - "Give me the full text of the matches" -> include_text=true

Notes:
  - Narrow with `poet_id` or `cat_id` first: unfiltered search over all poems is broad and slow
    to page through.
  - Set `include_text=true` only for a small page; full poems are long.
- **ganjoor_semantic_search_poems** — Meaning-based ("find a poem about...") search over Ganjoor's poetry collection.

Use this when you want poems by theme or feeling rather than an exact word. Describe the idea in
natural language — in English or Persian — and Ganjoor's embedding search returns the closest
verses. Prefer ganjoor_search_poems when you have an exact phrase to look for.

Args:
  - query (string): Natural-language description of the poem's theme, 1-500 characters
  - top_k (number): How many poems to return, 1-50 (default: 10)
  - poet_id (number): Restrict to a poet id. 0 = all poets (default: 0)
  - cat_id (number): Restrict to a category id. 0 = all categories (default: 0)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, results: [{ id, title, full_title, full_url, plain_text }] } ordered by relevance.

Examples:
  - "Poems about patience and divine trust" -> query='patience and trust in God'
  - "غزلی دربارهٔ جدایی از معشوق" -> query='a ghazal about separation from the beloved'

Notes:
  - Ganjoor's semantic backend is an optional feature that some deployments leave disabled; when it
    is off this tool returns a 503 with a message saying so. In that case fall back to
    ganjoor_search_poems with a Persian keyword describing the theme.
  - Relevance ordering is approximate — always quote the returned text rather than paraphrasing it.
- **ganjoor_get_poem** — Fetch a complete poem by its Ganjoor URL, including verses, metre, rhyme, and recitations.

This is the main way to read a poem. The URL is the same as the browser address on ganjoor.net,
e.g. "hafez/ghazal/sh494". Use ganjoor_search_poems or ganjoor_list_category_poems to discover
URLs, and ganjoor_get_poem_by_id when you only have a numeric id.

Args:
  - url (string): Poem URL slug, with or without leading slash, e.g. 'hafez/ghazal/sh494'
  - include_category (boolean): Include the book/category metadata (default: true)
  - include_siblings (boolean): Include the list of other poems in the same book (default: false)
  - include_recitations (boolean): Include audio recitations (default: true)
  - include_images (boolean): Include manuscript images (default: false)
  - include_comments (boolean): Include published reader comments (default: false)
  - include_verse_details (boolean): Include per-verse ids, positions, and couplet summaries (default: true)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  Poem id, title, full title, poet, book/category path, plain text, and the ordered verse list grouped
  into couplets. For JSON: { id, title, full_title, full_url, poet, category, couplets_count, verses: [{ couplet_index, position, text }], recitations: [...], next_poem, previous_poem }

Examples:
  - "Show me Hafez's ghazal 494" -> url='hafez/ghazal/sh494'
  - "Read this poem with its commentary" -> include_comments=true

Notes:
  - Persian text is returned verbatim; it is right-to-left, so present it as-is.
  - Keep include_siblings and include_comments false unless needed — both can be very large.
- **ganjoor_get_poem_by_id** — Fetch a complete poem by its numeric Ganjoor id.

Same content as ganjoor_get_poem, but keyed on the numeric poem id instead of the URL. Use it when
you have an id from search results or another tool and do not want to build a URL.

Args:
  - poem_id (number): Numeric poem id
  - include_category (boolean): Include the book/category metadata (default: true)
  - include_recitations (boolean): Include audio recitations (default: false)
  - include_comments (boolean): Include published reader comments (default: false)
  - include_verse_details (boolean): Include per-verse ids and positions plus the prose summary for each couplet (default: true). Set false for a leaner payload; the verse text itself is always returned.
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  Identical shape to ganjoor_get_poem.

Examples:
  - "Show poem 2623" -> poem_id=2623
  - "What is poem 13836?" -> poem_id=13836

Notes:
  - Combined with ganjoor_get_page_url you can also resolve an id to its web address.
- **ganjoor_get_poem_verses** — Get just the verses of a poem by id — a lightweight alternative to fetching the whole poem.

Each verse comes back in order with its couplet index and whether it is the first or second hemi-
stich. Optionally restrict to a single couplet.

Args:
  - poem_id (number): Numeric poem id
  - couplet_index (number): Return only this couplet (0-based). Omit for the whole poem.
  - include_summary (boolean): Include the per-couplet prose summary (default: false)
  - include_original_text (boolean): Include the uncorrected original text where available (default: false)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, verse_count, verses: [{ id, order, couplet_index, position, text }] }

Examples:
  - "Quote the verses of poem 2623" -> poem_id=2623
  - "Show only couplet 3" -> poem_id=2623, couplet_index=3

Notes:
  - Position 0 is the first hemistich, position 1 the second; couplet_index groups them.
- **ganjoor_get_random_poem** — Get a random poem from Ganjoor, optionally from a specific poet.

Args:
  - poet_id (number): Poet to draw from. 0 = a random poet (default: 0). Hafez = 2.
  - include_recitations (boolean): Include audio recitations (default: false)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  A single complete poem in the same shape as ganjoor_get_poem_by_id.

Examples:
  - "Surprise me with a poem" -> no arguments
  - "A random Saadi poem" -> poet_id=7

Notes:
  - Non-deterministic: repeated calls return different poems.
- **ganjoor_get_hafez_faal** — Draw a random Hafez ghazal as a Persian divination (فال حافظ).

This is the traditional "ask Hafez" reading: one ghazal selected at random, to be interpreted in
the context of the question asked. Each call draws a different ghazal.

Args:
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  A complete Hafez ghazal in the same shape as ganjoor_get_poem_by_id.

Examples:
  - "Draw a faal for me" -> no arguments
  - "What does Hafez say about my question?" -> no arguments

Notes:
  - Non-deterministic. The interpretation is traditionally left to the reader.
- **ganjoor_get_page_url** — Resolve any Ganjoor page id (poem, category, or poet) to its public web URL path.

Useful when you have a bare numeric id from an index or a citation and need to cite or link it.

Args:
  - id (number): Any Ganjoor page id — poem, category, or poet
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { id, url, full_url }

Examples:
  - "What is the link for id 13836?" -> id=13836

Notes:
  - Ids are globally unique across poems, categories and poets on Ganjoor.
- **ganjoor_get_redirect_url** — Resolve a legacy or short Ganjoor URL to its current canonical address.

Ganjoor has renamed many sections over the years; this endpoint maps an old path to today's one.

Args:
  - url (string): The URL path to resolve, with or without a leading slash
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { requested_url, redirect_url, full_url, has_redirect } — redirect_url is null and
  has_redirect is false when Ganjoor has no redirect for that URL (which includes URLs that are
  already canonical).

Examples:
  - "Where does an old Hafez section URL go now?" -> url='hafez/divan/ghazal/sh1'

Notes:
  - Ganjoor only keeps explicit redirects for paths it actually renamed; most valid URLs return
    has_redirect=false rather than a target.
- **ganjoor_list_rhythms** — List every Persian prosodic metre (وزن) used in the Ganjoor corpus, with its verse-pattern notation.

Use this to translate a metre name into the id Ganjoor uses, or to discover the metre vocabulary
before calling ganjoor_find_similar_poems or ganjoor_analyze_poem_rhythm.

Args:
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 50)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, rhythms: [{ id, rhythm, verse_count }] }

Examples:
  - "What metres exist?" -> no arguments
  - "Find the metre 'فاعلاتن'" -> page through results looking for the pattern

Notes:
  - `rhythm` is the classical Persian scansion pattern; `verse_count` is how many Ganjoor poems use it.
  - There are roughly 50-70 metres; one page of 100 usually covers everything.
- **ganjoor_analyze_poem_rhythm** — Detect the prosodic metre (وزن) of a poem by its numeric id.

Runs Ganjoor's prosody engine over the poem's verses and returns the matched metre together with
its confidence and the competing alternatives considered.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, metre: { id, rhythm }, confidence, alternatives: [{ id, rhythm, score }] }

Examples:
  - "What metre is poem 2623 in?" -> poem_id=2623

Notes:
  - The detected metre may differ from the poem's stored metre for edited or composite texts.
- **ganjoor_analyze_poem_rhyme** — Detect the rhyme letter (قافیه) and rhyme pattern of a poem by its numeric id.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, rhyme, rhyme_letter, fail_verse } — `rhyme` is the full rhyming ending as the
  prosody engine sees it, and `rhyme_letter` is its last character.

Examples:
  - "What is the rhyme of poem 2623?" -> poem_id=2623

Notes:
  - Some poems (short texts, composite editions) have no determinable rhyme; those return
    rhyme=null plus a fail_verse note explaining where analysis stopped.
- **ganjoor_find_similar_poems** — Find poems that share a poem's prosodic metre and rhyme letter — the basis of Ganjoor's "related poems" (شعرهای مرتبط) pages.

Two Persian poems matching in metre and rhyme can be read as answers to each other (reply-poetry),
so this is also how you trace how one poet answered another.

Args:
  - metre (string): The prosodic metre pattern, exactly as listed by ganjoor_list_rhythms
  - rhyme (string): Optional rhyme letter filter (e.g. 'ا', 'ب'). Empty means any.
  - poet_id (number): Restrict to a poet id. 0 = all poets (default: 0)
  - cat_id (number): Restrict to a category id. 0 = all categories (default: 0)
  - couplet_count_min (number): Only poems with at least this many couplets (default: 0)
  - couplet_count_max (number): Only poems with at most this many couplets (default: 0, meaning no cap)
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Results per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, poems: [{ id, title, full_title, full_url, plain_text }] }

Examples:
  - "Which Hafez poems share metre 'فاعلاتن مفاعلن فعلن'?" -> metre='فاعلاتن مفاعلن فعلن', poet_id=2
  - "Poems in metre X rhyming on ب" -> metre='...', rhyme='ب'

Notes:
  - `metre` must match the metre notation exactly; call ganjoor_list_rhythms to get valid values.
  - A metre with no rhyme filter returns a very large set — always page.
- **ganjoor_get_poem_sections** — Get the structural sections of a poem — the stanzas (بند), couplets (مثنوی), and tercets (رباعی) it is built from.

Ganjoor models a poem as nested sections so that poetry written in distinct forms within one poem
(e.g. a ghazal that closes with a qet'e) can be analysed per-section.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, section_count, sections: [{ id, index, number, section_type, verse_type, metre, rhyme }] }

Examples:
  - "What is the structure of poem 2623?" -> poem_id=2623

Notes:
  - Section `index` values feed directly into ganjoor_get_related_sections.
- **ganjoor_get_section** — Get one section of a poem by its section id, including its verses and its own related sections.

Args:
  - section_id (number): Section id — get these from ganjoor_get_poem_sections
  - include_verses (boolean): Include the section's verses (default: true)
  - include_related (boolean): Include related sections from other poems (default: false)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  Section metadata, its metre/rhyme, optionally its verses and its related sections.

Examples:
  - "Show me section 38101" -> section_id=38101

Notes:
  - A section id differs from a poem id; use ganjoor_get_poem_sections first to map between them.
- **ganjoor_get_related_sections** — Find sections from other poems that match a given poem section in metre and rhyme.

This is the section-level counterpart of ganjoor_find_similar_poems and is how Ganjoor links
reply-poems and cross-poet allusions at stanza granularity.

Args:
  - poem_id (number): Numeric poem id
  - section_index (number): 0-based section index within the poem (from ganjoor_get_poem_sections)
  - skip (number): Results to skip (default: 0)
  - items_count (number): How many to return. 0 or negative returns all (default: 10)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, section_index, total, related_sections: [{ id, poem_id, section_index, poet_name, full_url }] }

Examples:
  - "What matches the first section of poem 2623?" -> poem_id=2623, section_index=0

Notes:
  - `items_count=0` returns every match, which can be thousands — prefer a bounded count.
- **ganjoor_get_couplet_sections** — Find which structural sections a specific couplet of a poem belongs to.

A verse carries up to four nested section indices; this endpoint resolves one couplet to the
stanzas that contain it, which you need when analysing a ghazal's opening or closing section.

Args:
  - poem_id (number): Numeric poem id
  - couplet_index (number): 0-based couplet number
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, couplet_index, sections: [{ id, index, number, section_type, metre, rhyme }] }

Examples:
  - "Which section is couplet 5 of poem 2623 in?" -> poem_id=2623, couplet_index=5

Notes:
  - An empty list means the couplet index is out of range for that poem.
- **ganjoor_get_language_tagged_sections** — List poem sections Ganjoor has tagged with a specific language, newest-scoped by poet.

Ganjoor tags translated or bilingual stanzas with a language code (for example `ar` for Arabic
quoted lines, or `en` for English ones). This finds those passages across the corpus.

Args:
  - language (string): Language code, e.g. 'ar' for Arabic, 'en' for English, 'fa-IR' for Persian
  - poet_id (number): Restrict to a poet id. 0 = all poets (default: 0)
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { language, count, page, page_size, has_more, next_page, sections: [{ id, poem_id, poem_title, poet_name }] }

Examples:
  - "Which passages quote Arabic?" -> language='ar'
  - "Arabic stanzas in Hafez" -> language='ar', poet_id=2

Notes:
  - Coverage is patchy: most Persian stanzas are untagged, so an empty result does not mean no
    such passage exists.
- **ganjoor_search_recitations** — Search the published audio recitations (روایت) of Ganjoor poems, optionally by reciter or poem.

Each recitation is a sung or spoken performance of one poem by a named artist, published on
ganjoor.net.

Args:
  - search_term (string): Optional text to match against the reciter's (artist's) name
  - poet_id (number): Restrict to a poet id. 0 = all poets (default: 0)
  - cat_id (number): Restrict to a category id. 0 = all categories (default: 0)
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, recitations: [{ id, poem_id, poem_title, poem_url, artist, title, publish_date, duration_seconds }] }

Examples:
  - "Any recitations by محمدرضا ضیاء?" -> search_term='محمدرضا ضیاء'
  - "Recitations of Hafez's ghazals" -> poet_id=2

Notes:
  - When scoped to a poet or category, results are ordered by poem id; otherwise by publish date,
    newest first.
- **ganjoor_get_recitation** — Get one published recitation by its numeric id, including its audio file location and poem reference.

Args:
  - recitation_id (number): Recitation id (from ganjoor_search_recitations or a poem's recitation list)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { id, poem_id, poem_title, poem_url, title, artist, artist_url, audio_url, publish_date, duration_seconds, review_status }

Examples:
  - "Details for recitation 37535" -> recitation_id=37535

Notes:
  - Audio URLs point at Ganjoor's CDN and are stable once published.
- **ganjoor_get_recitation_sync_info** — Get the per-verse timing map for a recitation, showing when each couplet begins in the audio.

Useful for following along with the text while a recitation plays.

Args:
  - recitation_id (number): Recitation id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { recitation_id, sync_point_count, total_duration_ms, sync_points: [{ verse_order, text, start_time_ms, start_time_seconds }] }

Examples:
  - "Sync data for recitation 37535" -> recitation_id=37535

Notes:
  - Not every recitation has timing data; an empty list is normal for older uploads.
- **ganjoor_get_poem_recitations** — List every audio recitation published for one poem.

Args:
  - poem_id (number): Numeric poem id
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, count, page, page_size, has_more, next_page, recitations: [{ id, artist, title, publish_date, duration_seconds }] }

Examples:
  - "How has poem 2623 been recited?" -> poem_id=2623

Notes:
  - Text fields are intentionally blank upstream; fetch the poem itself for its text.
- **ganjoor_get_category_top_recitations** — Get the most-liked recitations within a category — Ganjoor's "top 1" (بهترین) listings for a book.

Args:
  - cat_id (number): Category id
  - include_poem_text (boolean): Include each poem's text (default: false)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { cat_id, count, recitations: [{ recitation_id, poem_id, poem_title, artist, score, plain_text? }] }

Examples:
  - "The best-known recitations of Hafez's ghazals" -> find the ghazals category id, then call this

Notes:
  - Ordering is by community upvote score, which changes over time.
  - The ghazals category for Hafez is id 10.
- **ganjoor_list_people** — List the historical and literary figures Ganjoor has catalogued as related persons (شخصیت‌ها).

These are people who appear in a poem's commentary or annotations — rulers, prophets, companions,
poets' contemporaries — as distinct from poets themselves.

Args:
  - search (string): Optional filter — only people whose name contains this text
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { total, count, page, page_size, has_more, next_page, people: [{ id, name, birth_year_hijri, death_year_hijri, birth_location, death_location, importance, gender }] }

Examples:
  - "Who does Ganjoor know about?" -> no arguments
  - "Find the person named 'حافظ'" -> search='حافظ'

Notes:
  - Filtering is done client-side; the full list is a few hundred entries.
  - `importance` ranks how central the person is; higher means better documented.
- **ganjoor_get_person** — Get one person's full record — biography, dates, places, and external references.

Args:
  - person_id (number): Numeric person id (from ganjoor_list_people)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { id, name, description, wiki_url, birth_year_hijri, death_year_hijri, birth_location, death_location, family_tree_caption, machine_generated, importance, gender }

Examples:
  - "Who is person 22?" -> person_id=22

Notes:
  - `machine_generated` marks entries created automatically; their biographies may be sparse.
  - Follow with ganjoor_get_person_poems or ganjoor_get_person_relations for context.
- **ganjoor_get_person_poems** — Get every poem Ganjoor has tagged with a given person.

This is the strongest link between a historical figure and the poetry about them.

Args:
  - person_id (number): Numeric person id
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { person_id, total, count, page, page_size, has_more, next_page, poems: [{ id, title, full_url }] }

Examples:
  - "Which poems mention this figure?" -> person_id=<id>

Notes:
  - Tagging is curated and incomplete for older figures.
- **ganjoor_get_person_relations** — Get a person's kinship and affiliation edges — parents, siblings, spouses, teachers, and contemporaries.

Args:
  - person_id (number): Numeric person id
  - relation_type (string): Optional filter — only relations of this kind, e.g. 'father', 'son', 'spouse', 'student', 'teacher'
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { person_id, relation_count, relations: [{ id, from_person_id, from_person_name, to_person_id, to_person_name, relation_kind, relation_name }] }

Examples:
  - "Who was this person's teacher?" -> person_id=<id>, relation_type='teacher'

Notes:
  - Edges are stored undirected with a kind; the same edge appears once.
  - Use ganjoor_get_person_family_tree for the whole connected component.
- **ganjoor_get_person_family_tree** — Get the full connected kinship component reachable from a person — ancestors, descendants, spouses, and siblings.

This returns the whole clan cluster, not just one hop, which makes it the right tool for tracing a
figure's lineage across generations.

Args:
  - person_id (number): Numeric person id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { person_id, root_id, person_count, people: [{ id, name, birth_year_hijri, death_year_hijri, family_tree_caption }], relation_count, relations: [{ person1_id, person1_name, person2_id, person2_name, relation_label }] }

Examples:
  - "Show me this figure's lineage" -> person_id=<id>

Notes:
  - Components can be large for well-documented dynasties (the Kayumars tree has ~56 people).
  - Data quality varies: some edges are machine-generated and may be speculative — check the
    machine_generated flag per person.
- **ganjoor_get_recent_comments** — Get the most recent published reader comments across Ganjoor, optionally filtered.

Comments are scholarly notes attached to a specific couplet of a poem.

Args:
  - term (string): Optional text filter applied to comment bodies
  - filter_user_id (string): Optional UUID — only comments by this user
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, page, page_size, has_more, next_page, comments: [{ id, poem_id, poem_title, couplet_index, user, text, date, has_owner_response }] }

Examples:
  - "What are people discussing?" -> no arguments
  - "Comments about 'انتظار'" -> term='انتظار'

Notes:
  - Ordering is by recency, so results change over time — not suitable for questions whose answer
    must stay stable.
  - Replies are not included, only top-level comments.
- **ganjoor_list_faq_categories** — List the Ganjoor site's published FAQ (پرسش‌های متداول) categories.

These are editorial explanations of how Ganjoor works — text sourcing, version differences,
contribution rules.

Args:
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { count, categories: [{ id, title, description, item_count }] }

Examples:
  - "What topics does Ganjoor's FAQ cover?" -> no arguments

Notes:
  - Follow up with ganjoor_get_faq_items using a category id.
- **ganjoor_get_faq_items** — Get every published question-and-answer in a Ganjoor FAQ category.

Args:
  - cat_id (number): FAQ category id (from ganjoor_list_faq_categories)
  - page (number): Page number, starting from 1 (default: 1)
  - page_size (number): Items per page, 1-100 (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { cat_id, total, count, page, page_size, has_more, next_page, items: [{ id, question, answer, answer_excerpt }] }

Examples:
  - "How does Ganjoor decide which text to use?" -> cat_id=<id>, then read the answers

Notes:
  - Answers are Persian prose and may contain HTML markup from the source; return them verbatim.
  - Despite its name, answerExcerpt carries the complete body — fullAnswer is always empty upstream.
- **ganjoor_get_faq_item** — Get one Ganjoor FAQ entry by its numeric id.

Args:
  - faq_id (number): FAQ item id (from ganjoor_get_faq_items)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { id, category_id, question, answer, answer_excerpt, pinned }

Examples:
  - "Show FAQ item 7" -> faq_id=7

Notes:
  - Despite its name, answerExcerpt carries the complete body; fullAnswer is always empty upstream.
  - Only published items are served. An upstream 500 usually means the id does not exist — re-check
    the id from ganjoor_get_faq_items.
- **ganjoor_get_poem_images** — Get the manuscript and facsimile images Ganjoor holds for a poem.

These are photographs of the physical manuscripts, with a thumbnail and a full-resolution link per
image — useful for checking a textual variant against its source.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, image_count, images: [{ id, url, thumbnail_url, alt, description }] }

Examples:
  - "Show the manuscript for poem 2623" -> poem_id=2623

Notes:
  - Most poems have no images; an empty list is the normal case.
- **ganjoor_get_poem_quoted_poems** — Get the poems that a given poem quotes or alludes to, and the poems that quote it.

Matching in metre and rhyme is common in Persian poetry — poets answer each other by reusing the
other's metre and rhyme. This endpoint maps those links in both directions.

Args:
  - poem_id (number): Numeric poem id
  - only_both_poems (boolean): Only keep pairs where both poems claim the quotation (default: false)
  - published_only (boolean): Only keep published records (default: true)
  - main_list_only (boolean): Only keep records curated onto the main "related poems" list (default: false)
  - skip (number): Results to skip (default: 0)
  - items_count (number): How many to return. 0 or negative returns all (default: 20)
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, total, returned, quotes: [{ id, related_poem_id, related_poet_name, related_poem_title, related_poem_url, is_prior, quoted_couplet, quoting_couplet, claimed_by_both, note }] }

Examples:
  - "Which poems does this one answer?" -> poem_id=<id>

Notes:
  - Each record pairs the quoted couplet with the couplet in the other poem that quotes it, so the
    actual verbal borrowing is visible.
  - `is_prior: true` means this poem came first chronologically.
  - `items_count=0` returns everything, which can be large.
- **ganjoor_get_poem_effective_corrections** — Get the editorial corrections applied to a poem's text — the before/after record of every fix.

Useful for understanding why the text on Ganjoor differs from an older printed edition.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, correction_count, corrections: [{ id, verse_order, text, original_text }] }

Examples:
  - "What corrections were made to poem 2623?" -> poem_id=2623

Notes:
  - Each entry pairs the corrected text with the original, per verse order.
- **ganjoor_get_poem_geotags** — Get the geographic and historical locations tagged to a single poem.

A geo tag pins a poem to a place and to the Hijri/Gregorian date it commemorates — for instance a
poem about a battle fought at a named site.

Args:
  - poem_id (number): Numeric poem id
  - response_format ('markdown' | 'json'): Output format (default: 'markdown')

Returns:
  For JSON: { poem_id, tag_count, tags: [{ id, latitude, longitude, lunar_date, gregorian_date }] }

Examples:
  - "Where is the event in poem 2623?" -> poem_id=2623

Notes:
  - For a whole diwan, use ganjoor_get_category_geotags instead.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "ganjoor-mcp-server": {
    "command": "npx",
    "args": ["-y","ganjoor-mcp-server"]
  }
}
```

## Documentation & README

# [![GANJOOR MCP SERVER: MCP server for the Ganjoor (گنجور) Persian classical poetry API](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/.github/assets/banner.svg)](https://github.com/ganjoor)

<div align="center">

[![VERSION: 0.1.0](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/.github/assets/pills/version.svg)]([CHANGELOG.md](https://www.npmjs.com/package/ganjoor-mcp-server))
[![PLATFORM: Node >=18.17](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/.github/assets/pills/platform.svg)](package.json)
[![LICENSE: MIT](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/.github/assets/pills/license.svg)](LICENSE)
[![Ganjoor on mcprush](https://mcprush.com/badge/ganjoor-mcp/score.svg)](https://mcprush.com/to/ganjoor-mcp?ref=badge)


</div>

<div align="center">

**English** | [فارسی](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/README.fa.md)

</div>

**Ganjoor MCP Server** — the Model Context Protocol server for the Ganjoor Persian poetry API.

It lets an agent browse poets and collections, read poems with their verses, run prosody and rhyme
analysis, trace reply-poems across poets, look up recitations and commentary, and explore the
related-people and family-tree graphs.

**44 tools, read-only access. No API key required.**

---

## Contents

- [Install](#install)
- [Docker](#docker)
- [Configuration](#configuration)
- [Concepts](#concepts)
- [Tool reference](#tool-reference)
- [Response formats](#response-formats)
- [Usage](#usage)
- [Architecture](#architecture)
- [Development](#development)
- [Known limitations](#known-limitations)
- [Evaluation](#evaluation)
- [Contributing](#contributing)

---

## Install

Requires Node.js 18.17 or newer.

```bash
npm install ganjoor-mcp-server # for locally installed stdio transport. no need if using VS Code or Claude Desktop / Cursor.
```

### VS Code

> No need install npm package — the MCP server runs in-process and is launched by VS Code when needed.

Just click the badge below to install the Ganjoor MCP server in your VS Code workspace. It will
  automatically add the server to your `.vscode/mcp.json` config.

[![Install in VS Code](https://img.shields.io/badge/Install_in-VS_Code-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://vscode.dev/redirect/mcp/install?name=io.github.ehsanghaffar%2Fganjoor-mcp-server&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22ganjoor-mcp-server%22%5D%2C%22env%22%3A%7B%7D%7D)


Or add to `.vscode/mcp.json`:

```json
{
  "servers": {
    "io.github.ehsanghaffar/ganjoor-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "ganjoor-mcp-server"
      ],
      "env": {}
    }
  }
}
```

VS Code CLI:

```bash
code --add-mcp '{\"name\":\"io.github.ehsanghaffar/ganjoor-mcp-server\",\"command\":\"npx\",\"args\":[\"-y\",\"ganjoor-mcp-server\"],\"env\":{}}'
```

### Claude Desktop / Cursor

Just click the badge below to install the Ganjoor MCP server in your Claude Desktop or Cursor app. It will automatically add the server to your MCP config.

[![Install in Cursor](https://img.shields.io/badge/Install_in-Cursor-000000?style=flat-square&logoColor=white)](https://cursor.com/en/install-mcp?name=io.github.ehsanghaffar%2Fganjoor-mcp-server&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsImdhbmpvb3ItbWNwLXNlcnZlciJdLCJlbnYiOnt9fQ==)

Or add to your MCP config file:

```json
{
  "mcpServers": {
    "io.github.ehsanghaffar/ganjoor-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "ganjoor-mcp-server"
      ],
      "env": {},
      "type": "local",
      "tools": [
        "*"
      ]
    }
  }
}
```

### Other clients

<details>
<summary>Claude Code</summary>

Use the Claude Code CLI to add the io.github.ehsanghaffar/ganjoor-mcp-server MCP server:

```bash
claude mcp add io.github.ehsanghaffar/ganjoor-mcp-server npx -y ganjoor-mcp-server
```
</details>

<details>
<summary>Claude Desktop</summary>

Follow the MCP install [guide](https://modelcontextprotocol.io/quickstart/user), use the standard config above.
</details>

<details>
<summary>Codex</summary>

Create or edit the configuration file `~/.codex/config.toml` and add:

```toml
[mcp_servers.ganjoor-mcp-server]
command = "npx"
args = ["-y", "ganjoor-mcp-server"]
```

For more information, see the [Codex MCP documentation](https://github.com/openai/codex/blob/main/codex-rs/config.md#mcp_servers).
</details>

<details>
<summary>Gemini CLI</summary>

Follow the MCP install [guide](https://github.com/google-gemini/gemini-cli/blob/main/docs/tools/mcp-server.md#configure-the-mcp-server-in-settingsjson), use the standard config above.
</details>

<details>
<summary>OpenCode</summary>

Follow the MCP Servers [documentation](https://opencode.ai/docs/mcp-servers/). For example in `~/.config/opencode/opencode.json`:

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "io.github.ehsanghaffar/ganjoor-mcp-server": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "ganjoor-mcp-server"
      ],
      "enabled": true
    }
  }
}
```
</details>

<details>
<summary>GitHub Copilot CLI</summary>

GitHub Copilot CLI supports adding MCP servers interactively and through a config file.

#### Option 1: Interactive setup with `/mcp add`

1. Open GitHub Copilot CLI in interactive mode.
2. Run `/mcp add`.
3. Enter **Server Name**: `io.github.ehsanghaffar/ganjoor-mcp-server`.
4. Choose **Server Type**:
   - `STDIO` for command-based servers
   - `HTTP` for remote servers
5. Fill in command/url settings from the standard config above and set **Tools** to `*`.
6. Press `Ctrl+S` to save.

#### Option 2: Edit `~/.copilot/mcp-config.json`

```json
{
  "mcpServers": {
    "io.github.ehsanghaffar/ganjoor-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "ganjoor-mcp-server"
      ],
      "env": {},
      "type": "local",
      "tools": [
        "*"
      ]
    }
  }
}
```

For more information, see the [GitHub Copilot CLI MCP documentation](https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-mcp-servers).
</details>

<details>
<summary>GitHub Copilot Coding Agent</summary>

GitHub Copilot Coding Agent can use MCP servers to extend its capabilities. Use the configuration below specifically formatted for the Coding Agent:

```json
{
  "mcpServers": {
    "ganjoor-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "ganjoor-mcp-server"
      ],
      "env": {},
      "type": "local",
      "tools": [
        "*"
      ]
    }
  }
}
```

Add this configuration to your repository settings under **Copilot > Coding agent**. The `"tools": ["*"]` setting enables all available tools from the MCP server.

For more information, see the [GitHub Copilot Coding Agent MCP documentation](https://docs.github.com/en/copilot/how-tos/use-copilot-agents/coding-agent/extend-coding-agent-with-mcp).
</details>

### Install from this source

```bash
npm install
npm run build
npm start            # stdio transport (default)
```

#### VS Code

Add to `.vscode/mcp.json`:

```json
{
  "servers": {
    "ganjoor": {
      "type": "stdio",
      "command": "node",
      "args": ["${workspaceFolder}/dist/index.js"]
    }
  }
}
```

#### Claude Desktop / Cursor

Add to your MCP config file:

```json
{
  "mcpServers": {
    "ganjoor": {
      "command": "node",
      "args": ["/absolute/path/to/ganjoor-mcp/dist/index.js"]
    }
  }
}
```

### Streamable HTTP (remote / shared deployment)

```bash
HTTP=true PORT=3000 npm start
```

The MCP endpoint is `POST /mcp`, and `GET /health` reports server and upstream status. The HTTP
transport runs statelessly (a fresh server instance per request), so it scales horizontally. It binds
`127.0.0.1` by default and validates the `Origin` header to block DNS-rebinding attacks; set
`ALLOWED_ORIGINS` (comma-separated) if you need non-loopback origins.

### MCP Inspector

```bash
npm run inspector
```

---

## Docker

A multi-stage `Dockerfile` and a `docker-compose.yml` are included. The image is built on
`node:22-alpine` and contains only the compiled output and production dependencies.

```bash
docker compose up --build
```

The compose file starts the server in **streamable HTTP mode** on `http://localhost:3000/mcp`, with
`GET /health` reporting server and upstream status. `HOST` is set to `0.0.0.0` so the published port
is reachable from the host; every configuration variable from the table below works as an ordinary
environment variable. To run a single container without compose:

```bash
docker build -t ganjoor-mcp .
docker run --rm -p 3000:3000 -e HTTP=true -e HOST=0.0.0.0 ganjoor-mcp
```

### stdio transport over Docker

Clients that launch the server themselves (rather than connecting over HTTP) can use the `stdio`
profile, which opens stdin for the protocol stream and starts no HTTP listener:

```bash
docker compose --profile stdio run --rm ganjoor-mcp-stdio
```

### With Claude Desktop

To run the server through Docker instead of a local Node install:

```json
{
  "mcpServers": {
    "ganjoor": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "ganjoor-mcp"]
    }
  }
}
```

---

## Configuration

All settings are environment variables; none are required.

| Variable | Default | Purpose |
| --- | --- | --- |
| `HTTP` | *(unset)* | Set to `true` to serve streamable HTTP instead of stdio. |
| `PORT` | `3000` | HTTP port. |
| `HOST` | `127.0.0.1` | HTTP bind address. |
| `ALLOWED_ORIGINS` | *(empty)* | Extra allowed `Origin` values, comma-separated. |
| `GANJOOR_API_BASE_URL` | `https://api.ganjoor.net` | Upstream API base URL — point at a local RMuseum instance for development. |
| `GANJOOR_SITE_BASE_URL` | `https://ganjoor.net` | Front-end base used to build clickable links. |

---

## Concepts

Ganjoor's data model has three nested levels. Getting these straight is the difference between a
productive and a fruitless agent session.

```mermaid
graph TD
    P["Poet — e.g. Hafez, id 2, slug /hafez"] --> C1["Category — e.g. Ghazals, id 24"]
    P --> C2["Category — e.g. Qeta, id 25"]
    C1 --> M1["Poem — e.g. Ghazal 494, id 2623, /hafez/ghazal/sh494"]
    C2 --> M2["Poem"]
    M1 --> V["Verses (hemistichs, paired into couplets)"]
    M1 --> S["Sections (nested stanzas with metre + rhyme)"]
    M1 --> R["Recitations (audio)"]
    M1 --> CM["Comments (reader commentary)"]
    M1 --> Q["Quoted / related poems"]
```

- **Poet** — an author. Hafez is poet id `2`, Saadi `7`, Rumi `5`, Khayyam `3`, Ferdowsi `4`.
  `ganjoor_list_poets` resolves a name to an id.
- **Category** — a book or section inside a poet's work (CatType `0` general, `1` poems, `2` book).
  Hafez's root category is `9`; his ghazals are `24`.
- **Poem** — one text, identified by a numeric id and a URL slug such as `hafez/ghazal/sh494`.
- **Verse** — a hemistich. Two consecutive verses form a `couplet` (`versePosition` 0 and 1).

> **URL slugs need a leading slash.** The upstream API returns 404 for `?url=hafez/ghazal/sh494` but
> resolves `?url=/hafez/ghazal/sh494`. This server normalizes for you, so either form works.

> **Persian text is returned verbatim.** It is right-to-left with Arabic-script diacritics used for
   prosody. Do not transliterate or translate it unless the user explicitly asks.

---

## Tool reference

### Poets and collections

| Tool | Description |
| --- | --- |
| `ganjoor_list_poets` | All published poets with biography, dates, places. Paginated. |
| `ganjoor_get_poet` | One poet by id or slug, with biography and top-level books/sections. |
| `ganjoor_get_poets_by_century` | Every poet grouped into historical half-centuries. |
| `ganjoor_list_books` | Named collections, filterable by poet or name substring. |

### Categories

| Tool | Description |
| --- | --- |
| `ganjoor_get_category` | A category with breadcrumb, sub-sections, and contained poems. |
| `ganjoor_list_category_poems` | Paginated poems in a category, with optional title filter. |
| `ganjoor_get_category_geotags` | Geographic/historical places tagged under a category's subtree. |
| `ganjoor_get_category_people_graph` | People network behind a category's "characters" tab. |

### Poems

| Tool | Description |
| --- | --- |
| `ganjoor_search_poems` | Full-text search across poems, scoped by poet or category. |
| `ganjoor_semantic_search_poems` | Meaning-based search ("find a poem about…"). |
| `ganjoor_get_poem` | A complete poem by URL, with verses grouped into couplets. |
| `ganjoor_get_poem_by_id` | The same, by numeric id. |
| `ganjoor_get_poem_verses` | Just the verses — lighter than fetching the whole poem. |
| `ganjoor_get_random_poem` | A random poem, optionally from one poet. |
| `ganjoor_get_hafez_faal` | A random Hafez ghazal for Persian divination. |
| `ganjoor_get_page_url` | Resolve any page id (poem, category, poet) to its web URL. |
| `ganjoor_get_redirect_url` | Resolve a legacy Ganjoor URL to its current address. |

### Prosody, rhyme, and structure

| Tool | Description |
| --- | --- |
| `ganjoor_list_rhythms` | All Persian prosodic metres with usage counts. |
| `ganjoor_analyze_poem_rhythm` | Detect a poem's metre. |
| `ganjoor_analyze_poem_rhyme` | Detect a poem's rhyme letter and pattern. |
| `ganjoor_find_similar_poems` | Poems sharing a metre and rhyme — the reply-poem graph. |
| `ganjoor_get_poem_sections` | A poem's structural sections (bonds, masnavi bands, tercets). |
| `ganjoor_get_section` | One section with verses and related sections. |
| `ganjoor_get_related_sections` | Sections from other poems matching this one's metre and rhyme. |
| `ganjoor_get_couplet_sections` | Which sections a given couplet belongs to. |
| `ganjoor_get_language_tagged_sections` | Stanzas tagged with a non-Persian language code. |

### Audio recitations

| Tool | Description |
| --- | --- |
| `ganjoor_search_recitations` | Search recitations by reciter, poet, or category. |
| `ganjoor_get_recitation` | One recitation with its audio URL. |
| `ganjoor_get_recitation_sync_info` | Per-verse timing for following along with audio. |
| `ganjoor_get_poem_recitations` | Every recitation published for one poem. |
| `ganjoor_get_category_top_recitations` | Most-upvoted recitations within a category. |

### People

| Tool | Description |
| --- | --- |
| `ganjoor_list_people` | Historical and literary figures in Ganjoor's knowledge graph. |
| `ganjoor_get_person` | One person's biography, dates, and places. |
| `ganjoor_get_person_poems` | Poems tagged with a person. |
| `ganjoor_get_person_relations` | A person's kinship and affiliation edges. |
| `ganjoor_get_person_family_tree` | The whole connected kinship component from one person. |

### Commentary, FAQ, and supplementary material

| Tool | Description |
| --- | --- |
| `ganjoor_get_recent_comments` | Recent published reader commentary. |
| `ganjoor_list_faq_categories` | Ganjoor's editorial FAQ categories. |
| `ganjoor_get_faq_items` | All Q&A entries in one FAQ category. |
| `ganjoor_get_faq_item` | A single FAQ entry by id. |
| `ganjoor_get_poem_images` | Manuscript and facsimile images. |
| `ganjoor_get_poem_quoted_poems` | Poems this one quotes, and poems quoting it. |
| `ganjoor_get_poem_effective_corrections` | The editorial before/after record of text corrections. |
| `ganjoor_get_poem_geotags` | Places and commemorative dates tagged to one poem. |

Every tool is annotated `readOnlyHint`, `idempotentHint`, and `openWorldHint`, and none of them
modify state on Ganjoor.

---

## Response formats

Every tool takes a `response_format` argument:

- **`markdown`** (default) — human-readable output with headers, lists, and clickable links.
- **`json`** — the same data as a structured object, also returned as MCP `structuredContent` so
  clients that support it can consume fields directly without parsing text.

List tools share a consistent pagination envelope:

```json
{
  "total": 842,
  "count": 20,
  "page": 1,
  "page_size": 20,
  "has_more": true,
  "next_page": 2,
  "poems": [{ "id": 2623, "title": "غزل شمارهٔ ۴۹۴", "full_url": "/hafez/ghazal/sh494" }]
}
```

`page` and `page_size` are 1-based; `page_size` caps at 100. Responses over 25,000 characters are
truncated with an explicit `truncated` flag rather than silently cut.

Two upstream endpoints (`/api/ganjoor/poets` and `/api/ganjoor/rhythms`) ignore their paging
parameters and always return the whole collection. The server detects that and slices the requested
window locally, so `page` and `has_more` stay correct and consistent across every list tool.

Errors are returned as tool results (not protocol errors) with a message that names the endpoint and
suggests a next step — for example, a 404 points you at `ganjoor_list_poets` to find a valid id.

---

## Usage

**"Who is Hafez?"**

```
ganjoor_get_poet { "url": "hafez" }
```

**"Show me Hafez's ghazal 494"**

```
ganjoor_get_poem { "url": "hafez/ghazal/sh494", "include_recitations": true }
```

**"Which Hafez poems contain the word صبر?"**

```
ganjoor_search_poems { "term": "صبر", "poet_id": 2, "page_size": 10 }
```

**"What metre is this poem in, and who else wrote in that metre?"**

```
ganjoor_analyze_poem_rhythm { "poem_id": 2623 }
ganjoor_list_rhythms { "page_size": 100 }
ganjoor_find_similar_poems { "metre": "<rhythm from the previous step>", "page_size": 10 }
```

**"Who are this figure's ancestors?"**

```
ganjoor_get_person { "person_id": 22 }
ganjoor_get_person_family_tree { "person_id": 22 }
```

**"What does Hafez say about my question?"**

```
ganjoor_get_hafez_faal { }
```

---

## Architecture

```
src/
├── index.ts            # Process entry: stdio or streamable HTTP transport
├── server.ts           # createServer() — registers every tool
├── client.ts           # fetch wrapper, error mapping, pagination helpers
├── types.ts            # TypeScript interfaces for Ganjoor's payloads
├── constants.ts        # Base URLs, limits, server identity
├── schemas.ts          # Shared Zod schemas (pagination, URL slugs, ids)
├── format.ts           # Markdown rendering and formatting helpers
└── tools/
    ├── registry.ts     # defineTool() wrapper: consistent errors + both formats
    ├── poets.ts        # Poet discovery
    ├── categories.ts   # Category browsing
    ├── poems.ts        # Poem retrieval, search, URL resolution
    ├── analysis.ts     # Prosody, rhyme, sections
    ├── audio.ts        # Recitations
    ├── content.ts      # Images, quotations, corrections, geo tags
    └── people.ts       # People graph, comments, FAQ
```

Two conventions run through the codebase:

- **`defineTool()` wraps every registration.** It guarantees each tool reports failures as `isError`
  results with actionable messages, renders markdown or JSON consistently, and returns
  `structuredContent` alongside text.
- **`md()` builds documents.** It preserves explicit blank lines so markdown headings and lists are
  separated properly, while dropping `null`/`false` so conditional fragments disappear cleanly.

Upstream quirks are normalized in `client.ts` and the tool modules rather than leaking to callers —
for example the `{ poet, cat }` envelope on category endpoints, the nested `category.cat` on poem
payloads, and the leading-slash requirement on URL parameters.

---

## Development

```bash
npm run build       # compile to dist/
npm run typecheck   # type-check without emitting
npm test            # live smoke test against api.ganjoor.net
npm run inspect     # print real tool output for review
npm run clean
```

`npm test` boots the server over an in-memory transport, asserts every tool has a substantial
description, and invokes all 44 tools against the live API, failing on any error response.

---

## Known limitations

- **Semantic search may be disabled.** Ganjoor's embedding backend is optional; when a deployment has
  it turned off, `ganjoor_semantic_search_poems` returns a 503. The tool surfaces a clear message and
  points you at `ganjoor_search_poems` with a Persian keyword instead.
- **Redirect resolution is sparse.** Ganjoor only stores redirects for paths it actually renamed, so
  most valid URLs return `has_redirect: false` rather than a target.
- **Some tools always need authentication.** Endpoints behind a Ganjoor account (bookmarks, posting
  comments, uploading corrections) are out of scope by design — this server is read-only and
  anonymous.
- **The book catalog is curated.** `ganjoor_list_books` covers ~168 named collections and omits some
  major poets — Hafez has no entries. Use `ganjoor_get_poet` with `include_cats: true` instead, which
  reads each poet's sections straight from their root category.
- **No total counts on paged endpoints.** The upstream API pages without reporting totals, so
  `total` is omitted where unknown; `has_more` is inferred by fetching one row beyond the page.
- **Poetry text coverage is curated.** Person tagging, geo tags, and language tags are incomplete by
  nature — an empty result is usually a coverage gap, not proof of absence.
- **Date fields are sometimes approximate.** Ganjoor flags dates it is unsure about
  (`validBirthDate` / `validDeathDate`); the tools expose those flags so callers can filter.

---

## Evaluation

`evaluation/evaluation.xml` holds 10 read-only questions with verified answers. Each one requires
several tool calls to answer and was checked against the live API before being written down:

```bash
npx tsx tests/verify-evaluation.ts   # replays every question through the MCP tools
```

The questions deliberately avoid single-lookup trivia. They cross-reference poets against centuries,
walk the kinship graph, compare the prosody of a poem against the corpus-wide metre list, and resolve
a reply-poem chain through quotation records.

---

## Contributing

See [CONTRIBUTING.md](https://github.com/ganjoor/ganjoor-mcp-server/blob/HEAD/CONTRIBUTING.md).