# Cody AI MCP for AI Agents MCP > Cody AI connects your agent to a knowledge base trained on proprietary documents and web content. It lets your AI client manage specialized bots, import new company guidelines from URLs or files, and start conversations using only your private information. ## Overview - **Category:** customer-support - **Price:** Free - **Tags:** rag, ai-assistant, knowledge-base, chatbot-training, document-query, automated-support ## Description When customer questions are scattered across wikis, PDFs, and old support tickets, keeping an AI agent accurate is a nightmare. This MCP solves that by connecting to Cody AI, turning your internal documentation into actionable knowledge for your agents. You feed the system your corporate manuals, web pages, and policy documents, and it creates specialized bots trained only on that data. Your agent doesn't guess; it pulls precise answers directly from your source material. Through Vinkius, you connect this capability to any compatible AI client—whether it’s Claude or Cursor. You can manage the entire lifecycle of your knowledge base right through natural conversation. Need to add a new HR policy? Simply import that web page and let the system handle the rest. The bottom line is, your team gets instant, accurate answers from company data without ever needing manual API calls. ## Tools ### create_conversation Starts a fresh conversation session with a specific bot for immediate Q&A. ### get_bot_details Retrieves comprehensive details about an existing knowledge bot, including its training scope and configuration. ### get_document_status Checks if the AI has finished processing a document so you know exactly when it's ready for questions. ### import_webpage Takes content from a specific URL and adds it to your knowledge base, expanding a bot's data set. ### list_bots Shows every single specialized bot currently active in the Cody AI account. ### list_conversations Gives you a list of your most recent chat sessions for quick review. ### list_documents Retrieves an organized list of all documents currently stored in the knowledge base. ### list_folders Shows the folder structure of your entire knowledge base, helping you organize content logically. ### list_messages Retrieves the full message history for a specific conversation thread. ### send_message Sends a direct prompt or question to the AI within an active conversation. ## Prompt Examples **Prompt:** ``` List all the specialized bots we have right now. ``` **Response:** ``` **Knowledge Base Bots Overview** You currently run 4 active knowledge bots: * 📚 **HR Policy Bot:** Trained on Employee Handbook (2023) and Benefits Guide. * ⚙️ **Product FAQ Bot:** Focuses on setup guides and troubleshooting. * 💰 **Billing Bot:** Handles invoice questions and payment plans. * 📜 **Legal Compliance Bot:** Uses the latest regulatory documents. Which bot should we start a conversation with? ``` **Prompt:** ``` I found this new article about our updated privacy policy, can you add it to the Legal Bot? ``` **Response:** ``` ✅ Upload initiated for 'new_privacy_policy.html'. The system is now processing the content and indexing new vectors. The bot will be fully ready to answer questions regarding data retention in approximately 3 minutes. *Source files:* `legal_dept_folder/` *Current status:* Indexing (75% complete) ``` **Prompt:** ``` Show me the full chat history for the Product FAQ Bot regarding Widget X. ``` **Response:** ``` **Conversation History: Product FAQ Bot** | Date | User Query | Bot Response Summary | | :--- | :--- | :--- | | 2024-07-15 | How do I reset my widget? | Go to Settings > Profile. (Source: *Setup Guide v2*) | 2024-07-16 | What's the max throughput? | The limit is 50 units/hour. (Source: *Technical Spec Sheet*) | Need to send a new message or start a new topic? ``` ## Capabilities ### Train and List Knowledge Bots Retrieve a list of all configured bots and fetch detailed information about any specific bot to check its current training status. ### Manage Knowledge Sources Import content from external URLs or retrieve lists of existing folders and documents within your knowledge base structure. ### Monitor Document Syncing Check the syncing status of an imported document to confirm that the AI has finished processing it and is ready to answer questions based on its contents. ## Use Cases ### Onboarding a new product line. A Product Manager needs the AI to answer niche questions about a new feature. Instead of manually uploading dozens of PDFs, they use `import_webpage` to feed all the launch site documentation into a bot's knowledge base, making it instantly available for agent querying. ### Handling complex compliance inquiries. A support agent gets asked about conflicting policies. They ask their AI client to check multiple bots and use `list_messages` to review the full chat history, ensuring they cite the correct policy document from the knowledge base. ### Creating a dedicated department bot. A Knowledge Manager wants an HR bot that only talks about employee guidelines. They first use `list_bots` to confirm their current setup, then create and configure a new bot using `get_bot_details`, restricting its scope to the relevant folders. ### Troubleshooting outdated guides. A team notices an old support bot is referencing deprecated procedures. They use `list_documents` to find the original source file and then run `get_document_status` on the replacement guide, ensuring the new data has successfully replaced the old knowledge. ## Benefits - Stop copy-pasting answers. Use the `send_message` tool to query a bot directly against your company's entire document library, giving instant, source-cited answers. - Never start from scratch. Run `list_bots` to see every available specialized agent—from HR policy bots to product FAQ bots—in one view. - Keep documentation current. Use `import_webpage` to feed new guidelines or policies directly into the knowledge base via a simple URL, keeping your agents up-to-date automatically. - Stay organized with `list_folders`. You can see exactly where every document lives and group related content before training a bot on it. - Know when data is ready. Use `get_document_status` to confirm that newly uploaded files have finished syncing, so your agent doesn't answer questions based on incomplete knowledge. ## How It Works The bottom line is that you treat complex data management—like importing a web page or checking document status—just like chatting with your agent. 1. Add the Cody AI integration to your preferred AI client's toolset. 2. Provide your API Key from the Cody AI dashboard. Your agent now recognizes the knowledge base connection. 3. Use natural language prompts (e.g., 'Show me all bots') to manage bot configurations, import content, and start conversations. ## Frequently Asked Questions **How does Cody AI MCP help me update my company policies?** You can keep your knowledge base current by using the import features. You simply provide a URL, and the system automatically pulls that web page content into the correct bot's training data so it has the latest information. **Is Cody AI MCP better than just uploading PDFs to my chatbot?** Yes. This MCP lets you manage the whole process, including knowing when a document is ready (`get_document_status`). You can also organize content into specific bots for different topics, preventing confusion from mixed documents. **Can I use Cody AI MCP to train a bot on my internal wikis?** Yes. If your wiki pages are available online via a URL, you can feed them directly using the `import_webpage` tool. This is much faster and more reliable than manual data handling. **What if I need an agent to talk about multiple topics?** You should create separate bots for different topics (like HR, Billing, etc.). The MCP helps you manage these distinct agents using `list_bots` and ensures each one only uses the data relevant to its purpose. **Does Cody AI MCP handle document organization?** It does. You can use tools like `list_folders` to see your knowledge base structure, ensuring that when you add new content, it lands in the right place before training a bot on it.