Poe MCP. Manage AI Bots and Chain Model Responses from Code.

You're dealing with Poe—you need your AI agent to talk to it like it's just another service, not something you have to log into via a web dashboard. This server lets you manage and execute against all those bots programmatically. You don't wanna manually mess around on the site; you want your code to handle the whole thing.

Managing Your Bot Infrastructure

You can control the entire lifecycle of any Poe API bot using specific functions. Need a new bot? Use create_bot and specify all the parameters—the name, the model type, and the base URL—to get it set up right in your code. If you've got an existing bot that needs tweaks, run update_bot; it applies those changes immediately for any conversations that follow.

When a bot is dead weight or just obsolete, use delete_bot to pull it off the platform completely; remember, though, this action wipes out all history and settings, so you gotta be careful.

To check what's already running, run list_bots, which gives you a complete list of every bot under your connected Poe account. If you only care about one specific setup, use get_bot with the ID to pull up its detailed configuration and metadata for a quick review. You can also peep at all the publicly available bots on the platform that your own systems can connect to using list_available_bots.

Running Queries and Chaining Models

This is where the real power is. Forget running queries manually; you'll use query_bot to run a specified bot on Poe, allowing you to pull its response directly into your workflow. This lets you chain responses—say, getting GPT-4 to summarize a document, and then feeding that summary right into Claude for refinement or tone adjustment.

You can also test connectivity fast by running send_message, which simulates a user sending an actual message to the bot endpoint. If you want to just verify if a whole sequence of models works together, this function validates that connection point.

Monitoring and Debugging Data

You need hard data on how your bots are doing, right? You can pull usage statistics for any specified Poe bot using get_bot_stats. This gives you quantitative metrics like the total message count and the number of unique users who've actually interacted with it. If a bot starts acting weird or misses context, don't guess—use list_messages to fetch the recent conversation history.

It shows you every message content and the timestamps for serious debugging or if you just wanna analyze usage patterns.

The Bottom Line

In short, your agent uses this server to handle everything: it builds bots using create_bot, modifies them with update_bot or removes 'em with delete_bot; it runs complex logic by chaining models through query_bot; and you can track the whole performance picture—from message counts via get_bot_stats to specific conversation logs using list_messages.

You've got total programmatic control over your Poe chatbot infrastructure.