# MonkeyLearn MCP > MonkeyLearn analyzes raw text data via custom machine learning models. It classifies sentiment (positive, negative, mixed), pulls structured entities like names, dates, and organizations, and executes multi-step NLP workflows on user feedback or support tickets. ## Overview - **Category:** customer-support - **Price:** Free - **Tags:** text-classification, entity-extraction, sentiment-analysis, nlp, machine-learning, data-labeling ## Description Yo, listen up. This server connects your agent straight to MonkeyLearn's deep text analysis engine. You don't just get simple tags; you run custom machine learning models on raw user feedback or support tickets. It handles serious Natural Language Processing (NLP) tasks right through conversation. **Classify Text Sentiment & Topic:** To figure out the vibe of a chunk of text, you use `classify_text`, which classifies it by determining its sentiment, topic, or intent using your custom models. You can check what classification types are available by calling `list_classifiers`. If you wanna know more about any specific classifier model you're running, `get_classifier_details` gives you the whole rundown. Wanna see what tags are attached to those classifiers? Use `list_classifier_tags`. **Pull Structured Entities:** Need names, dates, or company logos pulled out of a messy paragraph? You use `extract_text_entities`. This tool pulls structured data—like people's names, organizations, and addresses—and spits it back to you in JSON format. To see which types of entity pullers you got, run `list_extractors`. If you wanna check the specs on a specific extractor, `get_extractor_details` gives you the full usage guide. You can also peep available tags for extractors with `list_extractor_tags`. **Execute Complex Workflows:** Don't want to call five different tools? Use `run_workflow`. This tool executes pre-defined, multi-step NLP pipelines—like first classifying a topic and then pulling all the names mentioned—all in one shot. You can see what complex processes you built up with `list_nlp_workflows`. **System Management & Status:** Before you start running anything, you gotta make sure your connection's good. Use `get_api_status` to confirm your MonkeyLearn account connectivity is live. To keep track of how models perform, `list_model_versions` lets you check historical versions for any specific model, so you can test older settings if the new ones are giving you trouble. Need details on a specific text classification or entity extraction tool? You've already seen `get_classifier_details` and `get_extractor_details`. This stack gives you everything you need to analyze raw data. ## Tools ### classify_text Classifies a given piece of text by determining its sentiment, topic, or intent against your custom models. ### extract_text_entities Pulls structured data like names, organizations, and dates from unstructured text input. ### get_api_status Checks the current status of your MonkeyLearn account to confirm API connectivity is active. ### get_classifier_details Retrieves detailed information about a specific text classification model you own or use. ### get_extractor_details Gets the full specifications and usage guide for a particular entity extraction tool. ### list_classifier_tags Lists all available tags or labels associated with your text classification models. ### list_classifiers Returns a list of all the sentiment, topic, and intent classifiers you have access to. ### list_extractor_tags Shows available tags for your entity extraction models. ### list_extractors Returns a list of all the entity extractors, like person or address pullers. ### list_model_versions Retrieves historical versions for specific models, allowing you to test older performance settings. ### list_nlp_workflows Lists the custom multi-step NLP pipelines (Workflows) you have built in MonkeyLearn Studio. ### run_workflow Executes a pre-defined, complex NLP workflow using multiple steps and tools on new text data. ## Prompt Examples **Prompt:** ``` Classify this customer review: 'The product is amazing but delivery was slow.' ``` **Response:** ``` Classification (Classifier: cl_pi3C7nuL): Tag: 'Mixed' (Confidence: 0.82). Sub-tags: 'Product Quality' → Positive (0.95), 'Delivery' → Negative (0.88). Full text analyzed successfully. ``` **Prompt:** ``` Extract entities from: 'John Smith from Apple Inc. visited our NYC office on March 15.' ``` **Response:** ``` Extraction (Extractor: ex_YCya9nrn): Person: 'John Smith' (Confidence: 0.97). Organization: 'Apple Inc.' (0.95). Location: 'NYC' (0.92). Date: 'March 15' (0.99). Total entities found: 4. ``` **Prompt:** ``` List all my classifiers and extractors. ``` **Response:** ``` Classifiers: 3. 1) 'Sentiment' (cl_pi3C7nuL, Custom, 500 samples). 2) 'Topic Detection' (cl_abc123, Pre-built). 3) 'Intent' (cl_xyz789, Custom, 1200 samples). Extractors: 2. 1) 'Entity Extractor' (ex_YCya9nrn, Pre-built). 2) 'Keyword Extractor' (ex_def456, Custom). Workflows: 1. ``` ## Capabilities ### Classify Text Sentiment & Topic Determine the emotional tone or subject matter of a block of text using trained classifiers. ### Pull Structured Entities Extract specific data points (names, locations, dates) from raw text into structured JSON format. ### Execute Complex Workflows Run multi-step NLP pipelines—for example, first classifying the topic, then extracting relevant entities—all in one call. ### Manage Model Metadata Retrieve details on available classifiers, extractors, and model versions to understand your data capabilities. ## Use Cases ### Triage Incoming Support Tickets A support team gets 50 tickets in an hour. Instead of having agents manually read each one, the agent sends all text to `classify_text`. It immediately filters the list, telling you which are 'Billing' (Topic) and 'Angry' (Sentiment), allowing priority dispatching. ### Processing Survey Data at Scale Marketing receives 10,000 survey responses. Instead of a spreadsheet analysis, the agent loops through them, using `extract_text_entities` to pull out every mentioned competitor's name and product SKU, giving immediate competitive intelligence. ### Automating Research Summaries A researcher needs to analyze multiple articles. The agent uses a custom workflow (`run_workflow`) that first summarizes the text, then runs `classify_text` on the summary for 'Bias', and finally extracts key figures using `extract_text_entities`. ### Auditing API Functionality A developer needs to ensure their agent can run a complex pipeline. They first use `list_nlp_workflows` to see what's available, then call `get_api_status` to confirm the server connection is active before testing. ## Benefits - **Analyze Feedback Instantly:** Instead of manually reading hundreds of support tickets, use `classify_text` to get an immediate breakdown of sentiment (Positive/Negative) and the core topic. You'll know which issue needs attention before a human even looks at it. - **Structure Messy Data:** Stop treating key details like names or account numbers as messy text blobs. Use `extract_text_entities` to pull them into clean, structured JSON objects, ready for database insertion. - **Build Complex Logic in One Call:** Don't chain 5 different API calls. With `run_workflow`, you define a multi-step process—like 'Check Topic -> Extract Key Names -> Classify Urgency'—and get the final result in one go. - **Audit Your Tools Easily:** Before writing anything, use `list_classifiers` and `get_extractor_details`. This gives your agent the necessary context to know exactly which models exist for the job, preventing failed calls. - **Keep Track of Changes:** Use `list_model_versions` when a classifier updates. You can test older versions side-by-side with new ones, ensuring a model change doesn't break your core business logic. ## How It Works The bottom line is: you feed it raw text, and it gives you clean, actionable insights. 1. Subscribe to the server and input your MonkeyLearn API Key. 2. Your AI client sends text and a required tool call (e.g., `classify_text`) to the MCP Server. 3. The server runs the model, processes the data, and returns structured findings (tags, entities, or classifications) directly to your agent. ## Frequently Asked Questions **How do I check if the MonkeyLearn MCP Server is connected?** You call `get_api_status`. This tool quickly verifies your API key and connection status, confirming that the server can actually talk to your account before you run any heavy analysis. **Should I use classify_text or run_workflow for topic detection?** If topic detection is one of several steps in a larger process (e.g., Topic -> Entity Extraction -> Sentiment), use `run_workflow`. If it's the only thing you need, `classify_text` is simpler and faster. **What if I want to pull names from an address block?** Use `extract_text_entities`. This tool is designed specifically to isolate structured data like Persons, Organizations, or Locations even when they appear mixed into a paragraph of text. **How do I see what classifiers are available?** You call `list_classifiers`. This gives you an index of all the topic and sentiment models you can use without having to guess which one is correct or active. **What does the `get_api_status` tool confirm about my account access?** It verifies that your API key is active and correctly connected. This check confirms general service connectivity before you run intensive jobs, ensuring a quick diagnosis if there are authentication issues. **If I need to revert or debug, how does `list_model_versions` help me find an older classifier or extractor?** This tool retrieves all historical and current versions of your models. It's essential for debugging when a new deployment breaks analysis because you can pinpoint and select the last known working version ID. **What is the operational difference between calling `classify_text` versus using `run_workflow`?** `classify_text` executes a single, defined model check (like sentiment analysis). In contrast, `run_workflow` triggers complex pipelines built in Studio, allowing data to pass through multiple sequential actions automatically. **How do I see all predefined tags using the `list_classifier_tags` tool?** The tool retrieves every available tag associated with a specific classifier model. This lets you confirm exactly what labels your trained model recognizes, which is useful before initiating any text analysis jobs. **Can I classify text by sentiment or topic?** Yes. Point to any classifier model ID and pass text to get classification results with confidence scores. **How does MonkeyLearn authentication work?** MonkeyLearn uses `Authorization: Token {API_KEY}` header against `api.monkeylearn.com/v3`. **Can I extract named entities from text?** Yes. Use an extractor model to pull keywords, people names, organizations, locations, and more from raw text.