search_hybridLABSCO SUMMARY
Every answer comes with a link that proves it.
The failure mode with scripture and language models is a quotation that sounds right and is not in the text. Here each response carries a deep link to a reader that renders the source data verbatim with the cited segment highlighted, so verification is one click rather than an act of faith. The pairing of dictionary lookup with canonical self-definition is the other unusual bit — what a term means to a lexicographer and what the suttas say it means are different answers.
What it is
A free public server over the Pāli Tipiṭaka at parity with SuttaCentral — Sutta, Vinaya and Abhidhamma, in Pāli with English where an English translation exists upstream. Search combines keyword and semantic retrieval, and every response carries a deep link into a bilingual reader so the quoted line can be checked.
What you get
- search_hybrid — keyword and semantic results fused, the one to reach for on a concept
- search_by_keyword for an exact Pāli word with fuzzy matching, and search_semantic for pure similarity
- survey_corpus — an exhaustive count and per-basket breakdown, for 'how often and where does this appear' rather than best matches
- get_sutta by id, whole or as an outline, a segment range, or a window around a hit
- open_sutta_viewer, which renders the sutta inline as Pāli and English side by side with the cited segment highlighted
- get_reference for a properly formatted academic citation with its source URLs
- compare_translations across editions, aligned at the segment level
- get_word_definition against PTS, DPPN and the Payutto dictionary, and parse_pali_word to recover a root form when a lookup misses
- define_from_suttas — how the canon defines a term in its own words, rather than how a dictionary glosses it
- list_structure and list_editions for coverage and what is loaded
Requirements
Nothing: the maintainers run a free public instance and offer it as Dhamma Dāna, non-commercial only. Streamable HTTP is the modern endpoint, with legacy SSE alongside it; older clients bridge through mcp-remote, which needs Node. Self-hosting is Python with PostgreSQL and pgvector, or a fully offline local build. Abhidhamma is Pāli only, because no English translation exists upstream.
Setup effort
One command — pipx install tripitaka-mcp
USE CASES
01
Finding what the canon says about a topic
Search across the corpus with keyword and meaning together, then open the passage that matched.
02
Quoting a passage you can stand behind
Fetch the sutta and generate the formatted citation with its source URLs.
03
Working out what a Pāli word means
Look it up in the dictionaries, and strip its inflection first when the lookup misses.
04
Asking how the texts define a term themselves
Find the canonical definition formulas rather than a modern gloss.
05
Checking whether translators disagree
Compare renderings of the same segment across editions, aligned line by line.
06
Answering how often something appears
Run the exhaustive survey for a real total and a per-basket breakdown, not the top few hits.
07
Reading a passage properly in context
Render it inline in Pāli and English with the cited segment highlighted.
13 tools
Listed in project docs. Your client may show a different list.
search_by_keywordsurvey_corpussearch_semanticget_suttaopen_sutta_viewerget_referencecompare_translationslist_structurelist_editionsget_word_definitiondefine_from_suttasparse_pali_wordBEFORE YOU INSTALL
What to know first
Tools you get
13 tools named in the project's documentation.
Requires
- Node.js with npx, for the hosted route through mcp-remote
- Python 3.10+ with pipx, for the offline local edition
INSTALLATION
Connect your client
Server setup
pipx
TERMINAL
pipx install tripitaka-mcpHosted endpoint
ENDPOINT URL
https://mcp.tripitaka-mcp.com/mcpAdd this hosted endpoint to your MCP client.
MCP Clients
Claude Desktop — remote
Find your absolute npx path
Claude Desktop does not read your shell profile, so a bare npx will not resolve.
CLAUDE DESKTOP — REMOTE · TERMINAL
which npxAdd this to claude_desktop_config.json
<YOUR_NPX_PATH> is the path which npx printed (for example /Users/you/.nvm/versions/node/v22.14.0/bin/npx), and <YOUR_NODE_BIN_DIR> is that path's parent directory. Then quit Claude Desktop completely and reopen it.
CLAUDE DESKTOP — REMOTE · JSON
{
"mcpServers": {
"tripitaka": {
"command": "<YOUR_NPX_PATH>",
"args": [
"-y",
"mcp-remote",
"https://mcp.tripitaka-mcp.com/mcp"
],
"env": {
"PATH": "<YOUR_NODE_BIN_DIR>:/usr/local/bin:/usr/bin:/bin"
}
}
}
}Claude Desktop — local, Cursor — local
JSON
{
"mcpServers": {
"tripitaka": {
"command": "tripitaka-mcp",
"args": [
"serve"
]
}
}
}Any MCP client — legacy SSE
ANY MCP CLIENT — LEGACY SSE · JSON
{
"mcpServers": {
"tripitaka-mcp": {
"url": "https://mcp.tripitaka-mcp.com/sse"
}
}
}
