Writer (AI Enterprise LLM) MCP. Query private knowledge graphs and generate content from internal data.

You wanna run serious AI stuff? This server hooks your agent up with enterprise-grade models and private data graphs, so you can do real work—not just some fluffy chatbot nonsense. You use this when you need to generate content using Palmyra models, build Knowledge Graphs from proprietary docs, or query specific internal data sets using RAG workflows.

Querying Internal Data & Knowledge Graphs
You've got private documents? Use create_graph to set up a new Knowledge Graph structure for your company knowledge. Then you upload files with upload_file, and connect them to that graph using add_file_to_graph. You can list all graphs created with list_graphs, and check out every single file stored on the server by calling list_files.

When you need answers, you ask a question using ask_question, which queries one or more Knowledge Graphs based only on your private data. If you gotta take a file out of the graph later, use remove_file_from_graph.

Generating Content and Text Completions
Need text? You can generate simple plain text completions using text_completion, which takes one explicit prompt. For full conversational responses or complex content creation, you'll use chat_completion. To get a list of available models (like palmyra-x5), run list_models. If your content job fails, don't panic; just try running it again with retry_application_job.

Handling Documents and Vision Analysis
Got PDFs or images? You can upload them using upload_file. To convert a PDF into readable text, use the deprecated but available parse_pdf tool. If you need to analyze visuals—say, an image or another document—based on what you ask, run analyze_vision. When you're done with a file, you can get its basic metadata (size, owner) using get_file, or permanently delete it from storage with delete_file.

Managing AI Applications and Workflows
This server lets you manage content generation jobs built into no-code applications. You can see what apps are ready to go by calling list_applications. To figure out how an app works, use get_application, which retrieves its configuration details. If you want to run a job immediately using a defined application, call generate_application_content.

If the job is big and takes time, start it asynchronously with generate_application_content_async and keep the resulting job ID handy. To check on that background process, use get_application_job; if you need to see all past jobs for an app, run list_application_jobs. You can also get a list of all deployed applications using list_applications.

System Utilities and Translation
For extra data handling, you can download the raw content of any file with download_file. If you need to change languages, use translate_text for translation between supported tongues. For public context, run web_search using a query string. Finally, if you have an existing application ID and just want to check its details without running it, you can still call get_application.