TextLite MCP Server
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MCP_HOST | No | Host for the MCP server. | 127.0.0.1 |
| MCP_PORT | No | Port for the MCP server. | 8000 |
| MCP_TRANSPORT | No | MCP transport to use. | stdio |
| TEXTLITE_MAX_TEXTS | No | Maximum number of input texts accepted by the server. | 100 |
| TEXTLITE_QUERY_MODEL | No | Query model used by TextLite. | sentence-transformers/multi-qa-MiniLM-L6-cos-v1 |
| TEXTLITE_RERANK_MODEL | No | Cross-encoder reranking model used by TextLite. | cross-encoder/ms-marco-MiniLM-L-6-v2 |
| TEXTLITE_EMBEDDING_MODEL | No | Embedding model used by TextLite. | sentence-transformers/all-MiniLM-L6-v2 |
| TEXTLITE_MAX_TOTAL_CHARS | No | Maximum total characters across input texts. | 120000 |
| TEXTLITE_MAX_CHARS_PER_TEXT | No | Maximum characters per input text. | 20000 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Semantic NLP tools for deduplication, clustering, context compression, summarization, keyphrases, topic extraction, reranking and search. Input sizes are capped. Models run locally and may download on first use.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
| prompts | {
"listChanged": true
} |
| resources | {
"subscribe": false,
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| server_infoB | Return server version, configured models and input limits (no models loaded). |
| deduplicate_textsC | Remove normalized-exact and semantically similar duplicates; preserve duplicate groups. |
| cluster_entitiesC | Cluster related entity names or phrases; output members, label, representative and cohesion. |
| compress_contextsC | Choose informative, diverse paragraphs under an approximate text-token budget. |
| compress_query_contextsC | Extract question-relevant sentences under an approximate text-token budget. |
| semantic_similarityB | Return the cosine similarity (-1 to 1) between two locally embedded texts. |
| semantic_searchC | Return top-k candidate texts ordered by cosine similarity to a query. |
| summarize_textB | Produce an extractive (not generative) summary with selected source sentences. |
| summarize_paragraphsC | Extractively summarize each paragraph; optionally apply a shared approximate token limit. |
| rerank_passagesB | Rerank passages with a cross-encoder; returns passages in relevance order (no raw scores). |
| extract_keyphrasesC | Extract scored, frequency-aware keyphrases with configurable n-gram size. |
| keyphrases_to_tagsC | Convert output from extract_keyphrases into annotation-style tag dictionaries. |
| extract_topicsC | Discover high-level topics by clustering extracted keyphrases. |
| assign_entities_to_topicsA | Attach entities by lightweight lexical matching to topics from extract_topics. topics entries need id, label, score, size, keyphrases. entities entries need name and optionally id, weight, importance_score. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 14 tools
Most tools have clearly distinct purposes, and adjacent tools like semantic_similarity, semantic_search, and rerank_passages are differentiated by their inputs and outputs. There is some potential overlap among clustering/topic tools and among compression/summarization variants, but descriptions generally clarify the boundaries.
Names are consistently snake_case and mostly follow a verb_noun pattern such as extract_keyphrases and deduplicate_texts. A few tools use noun or adjective phrases like semantic_search, server_info, and keyphrases_to_tags, which are minor deviations but still readable.
With 14 tools, the server is well-scoped for a text-processing toolkit. Each tool covers a recognizable operation, and the count is within the typical range for a coherent multi-function server.
The surface covers major text-lite workflows: deduplication, compression, summarization, semantic search/similarity, reranking, keyphrase extraction, topic extraction, and entity assignment. Some utilities such as explicit embedding generation or batch orchestration are absent, but core operations are well represented.