Heidelberg AICurriculum
Track 10 · Intermediate
10.1.3

Poe

Many models, one chat window — and bots anyone can build

7 lessons 2026-08-06 AI-generated

1Overview

Poe, from Quora, was the first mainstream way to talk to several frontier models from one interface, and it still has the shallowest learning curve of the platforms here. Alongside the official models it hosts user-built bots — a system prompt, sometimes a knowledge base, published for anyone to use. → Reach for it to compare how different models answer the same question without opening four accounts, or to publish a small assistant without writing an app. → Unlike Langdock or Abacus, it is aimed at individuals rather than at a company rolling out a workspace.

Not written yet — this chapter is a placeholder for Poe, one of the all-in-one AI platforms. Research and copy still to come.

1.1How many models can I access?

You can chat with several frontier AI models all within Poe’s single interface, letting you compare their responses without juggling multiple accounts.

1.2Can I create my own bot?

Yes—you can publish a custom bot by defining a system prompt or knowledge base, making it available for anyone to use on the platform.

1.3Is Poe suited for personal projects?

Poe is designed for individual users, offering an easy learning curve and tools that let you experiment with AI assistants without building a full application.

2Lessons 7

2.1 Compare answers from two models in Poe

The model chip in the compose bar lets you select which AI provider answers your question.

Ask the same question to two different models without leaving the page

  1. Open poe.com in a browser and sign in using Google, Apple or email
  2. Click the model chip in the compose bar to display the list of available models
  3. Select a model from the list and click the Send button to submit your question
  4. Click the model chip again, choose a different model, and click Send to resend the same question
  5. Observe the two replies appear as separate chat threads
  • You'll see Two separate replies appear side by side, showing differing length and confidence for the identical prompt
  • Takeaway Switching the model chip lets one account route a single query to multiple providers instantly
  • Check How does using the model chip let you see two different replies to the same prompt without leaving the page?
  • Cost Free at normal usage. A free account has daily limits on how many messages and which models you can reach; you will hit that ceiling if you experiment heavily, which lesson 02 explains in point terms.

2.2 Check a bot's live cost and plan your monthly points usage

The point value displayed next to the Send button shows the live cost of each reply, while the Points page lists your balance and reset date.

Determine how many points each reply will cost and size your subscription or top‑up to match observed usage

  1. Open a chat window and look at the point value displayed next to the Send button on the compose bar
  2. Click the Account icon, then select Points to view your current balance and the reset date for the billing period
  3. Choose Subscription plan from the Settings menu to see the monthly points allocation for each tier
  4. If needed, click Add points and select a One‑off top‑up amount to purchase additional points that expire in one year
  • You'll see The compose bar shows a point value next to the Send button and the Points page displays a decreasing total with the period reset date
  • Takeaway Points are shared across all models, each bot shows its live cost before you send, and plans reset monthly while top‑ups expire after a year
  • Check What information on the compose bar and Points page tells you how many points a reply will cost and when your monthly allocation resets?
  • Cost €4.79–€239.58/month depending on tier (checked 2026-08-08, EUR shown; confirm the localized price at signup), or ~$30 per 1M pre-purchased points as a top-up. A free account exists but is the most limited on both message volume and which models are reachable.

2.3 Choose the best model for your own task

The output from each model, shown as separate replies, lets you compare usability for your prompt.

Select a model that delivers usable output for a specific prompt you need

  1. Open the model picker via the Compose bar chip and note the current list of available models
  2. Run your chosen task (e.g., summarise a document, draft an email, sketch a function signature) through three different models from separate providers
  3. Observe the output of each model and compare usability for your specific need
  4. Record which model performed best for that type of task
  • You'll see Three replies to the same prompt, each showing different strengths and weaknesses
  • Takeaway The roster changes regularly; the lasting habit is to test multiple models on your own prompts rather than rely on benchmark scores
  • Check After testing three models on the same task, what should you record to decide which model is best for that specific need?
  • Cost Points per lesson 02's mechanics — running the same prompt three times costs roughly three times one model's per-message cost, since each is a separate full reply.

2.4 Create and publish a prompt bot

The Publish button makes the bot live at its own URL, and opening a chat verifies its replies match the system prompt.

Create, configure, and publish a working prompt bot — a system prompt plus optional knowledge — using Poe's in‑browser creator tools.

  1. Open the Poe sidebar or navigate to poe.com/creators and click Create bot.
  2. Enter a name, description, and write the system prompt in the provided fields.
  3. Upload reference files using the knowledge section if needed.
  4. Select the underlying model from the model picker control.
  5. Click Publish to make the bot live.
  6. Open a fresh chat with the bot to verify that its replies follow the system prompt.
  • You'll see A bot page at its own poe.com URL, listed under your profile's Creators tab, answering messages from a chat window identical to any official model's.
  • Takeaway A prompt bot is a system prompt plus optional attached knowledge, built in the browser and published to a real URL in minutes — no code, no hosting
  • Check Which steps confirm that a newly created bot follows its system prompt when you start a fresh chat with it?
  • Cost Free to create. Every message a user sends it still spends points from that user's own pool (or the creator's, depending on configuration) at the underlying model's rate.

2.5 Choose the correct bot type for your task

The bot type selector shows Prompt Bot for static answers and Server Bot for live data, API calls, or persistent state.

Decide whether a no‑code prompt bot or a hosted server bot is required before building.

  1. Assess whether the task needs live data, an external API call, or persistent state.
  2. If it does not, select Prompt Bot when creating a new bot.
  3. If it does, choose Server Bot and follow Poe’s creator documentation to configure your service.
  • You'll see The bot settings page shows either “Prompt Bot” or “Server Bot” selected, matching the capability you need.
  • Takeaway Use a prompt bot only for static answers from uploaded documents; any live lookup, external action or persistent state demands a server bot you host yourself
  • Check What determines whether you should choose a Prompt Bot or a Server Bot when setting up a new bot?
  • Cost A prompt bot: free to build, same per-message point cost as any bot. A server bot: your own hosting bill, on top of whatever the underlying model still charges through Poe.

2.6 Identify when Poe is unsuitable

The Data & Privacy tab and Points Usage panel reveal handling policies, cost per message, and lack of team admin features.

Spot situations where Poe cannot meet your requirements

  1. Open Account Settings from the user menu
  2. Select the Data & Privacy tab to view Poe's handling policies
  3. Read the linked Subscriber Terms and any data‑use policy displayed there
  4. Check the Points Usage panel to see how message volume translates into cost
  5. Verify whether a Workspace Admin Console is available – if it is absent, team management is not supported
  • You'll see The Data & Privacy page shows the provider terms that govern how your messages are handled
  • Takeaway Poe lacks self‑hosting, volume discounts and team admin features, so it fits solo model comparison but not sensitive data, heavy traffic or enterprise rollout
  • Check Which sections of the Account Settings help you decide if Poe’s policies and point costs make it unsuitable for your project?
  • Cost No direct point cost — this is a decision, not an action.

2.7 Choose Poe over other all‑in‑one AI platforms

The Model selector lets you switch providers mid‑conversation, while the Bot gallery displays community‑built bots for quick publishing.

Identify when Poe is the right choice instead of Langdock, Abacus.AI or Genspark and justify that decision

  1. Open the Poe home page in a browser.
  2. Select a different model from the Model selector to switch mid‑conversation.
  3. Browse the Bot gallery to find a bot created by another user.
  4. Compare the responses shown for each selected model.
  • You'll see Poe’s home screen shows a model selector for switching models and a bot gallery listing community‑built bots
  • Takeaway Pick Poe when an individual needs to compare raw model output or publish a small bot without team overhead, otherwise choose a workspace platform for shared documents and seat management
  • Check How do the model selector and bot gallery on Poe’s home screen help you justify choosing Poe over other all‑in‑one AI platforms?
  • Cost Free — this is a comparison read, no messages sent.

3You’ll know it worked 14 checkable outcomes in this chapter

  • You can open the bot’s public URL or find it in Explore and have it respond to your test messages
  • The image appears in the chat window and can be saved or upscaled within Poe
  • The output contains key sections (abstract, intro, results) without truncation errors
  • When you open the bot from your Poe dashboard and send a test message (e.g., "I am having a bad day"), it returns an empathetic response matching your character prompt.
  • The bot appears in the “Community Bots” list and responds per your system prompt when you chat with it
  • You see the bot appear in your chat window and it responds to your prompt
  • Your bot appears on the Explore page under your username and responds correctly when opened
  • The error disappears and you can send new messages in Poe

14 outcomes in all — one per recipe below.

4FAQ, Tips & How-to 14

one problem, one solution, one action
How-to Everyone

I want a custom role‑play chatbot without coding

The video walks through building a new bot on Poe by selecting a base model, defining its character, adding knowledge sources, adjusting temperature and other settings, then publishing it. This shows how to turn a simple idea into an interactive AI assistant without coding.

How to Digital ↗ Lesson → AI-generated
How-to Everyone

Want a picture from a simple text prompt

The tutorial demonstrates using Poe’s pre‑made image generation bots by entering a descriptive prompt, then upscaling or animating the result. This shows how to leverage Poe’s multimodal models for quick visual content without leaving the platform.

How to Digital ↗ Lesson → AI-generated
How-to Everyone

Need to test different AI answers side by side

Poe lets you select from many large language models (e.g., GPT‑4, Claude, PaLM 2) without leaving the interface. The left sidebar lists available models; clicking a model changes the backend engine for your current conversation.

Dylan Curious ↗ Lesson → AI-generated
How-to Everyone

Long article or research paper needs summarizing

Models like GPT‑4 32k or Claude 100K have huge context windows, allowing you to paste thousands of words and get a coherent summary in one request. This bypasses the 8K limit of standard ChatGPT.

Dylan Curious ↗ Lesson → AI-generated
How-to Everyone

I need a chatbot that always uses my own intro

Poe’s Bot Builder lets you define a permanent system prompt (or “base prompt”) that shapes how the assistant behaves, effectively giving you a lightweight, reusable AI without coding.

Dylan Curious ↗ Lesson → AI-generated
How-to Everyone

Need a caring chatbot but can’t code

The tutorial walks through Poe's no‑code bot builder: signing up, selecting the Role‑Play Bot type, naming it, writing a character prompt, adding optional knowledge sources and adjusting settings like temperature. By defining clear instructions in the character definition, you shape how the bot responds.

NIIT ↗ Lesson → AI-generated
How-to Everyone

Can't tell which AI assistant fits your task

The model library on Poe groups dozens of chat, image, video, audio, and code assistants by category, letting you see at a glance which model excels at a given task. By selecting different models in the same chat window you can instantly compare their responses without opening separate accounts.

AI Outlet ↗ Lesson → AI-generated
How-to Everyone

No coding needed for a FAQ helper

Poe lets any user build a custom bot by defining a system prompt and optional knowledge base, then publishing it for others to use. This provides a zero‑code way to create specialized assistants such as a FAQ helper or a writing coach.

AI Outlet ↗ Lesson → AI-generated
How-to Everyone

Need a ready‑made chatbot for a specific task

Poe’s Explore page lists thousands of community bots, each pre‑configured for a specific task. Browsing it lets you quickly find a bot that matches your need without building one yourself.

Pro Assistant ↗ Lesson → AI-generated
How-to Everyone

Want to see how different AI models answer the same question

Poe lets you change the underlying model of an active conversation, so you can ask the same question to multiple models and see their distinct reasoning styles side by side.

Pro Assistant ↗ Lesson → AI-generated
How-to Everyone

Need a shareable chatbot but don’t want to code

Poe allows you to turn a custom system prompt into a shareable bot, enabling anyone to use your assistant simply by clicking “Add to Chat.” No app development is required.

Pro Assistant ↗ Lesson → AI-generated
How-to Everyone

Want a personal assistant in your chat window

Poe lets you define a new bot by giving it a name and selecting the underlying model or tool (e.g., image generator, prompt helper). This creates a reusable chat endpoint that appears alongside official bots.

1ClickTutorialen ↗ Lesson → AI-generated
How-to Everyone

Reached my daily chat limit

When you exceed Poe’s daily point quota, the platform blocks further messages. You can either wait for the quota to reset (usually after 24 hours) or upgrade to a paid plan that provides more points per day.

1ClickTutorialen ↗ Lesson → AI-generated
Tip Everyone

Understanding Poe’s point system — estimate costs of different bots

Poe charges points based on token consumption: simple queries use few points, while long or complex requests (e.g., generating a 40‑page novel) consume many. Knowing this helps you budget daily usage within the free 300‑point allowance or your paid tier.

The same set on /recipes, filtered by tool and role.

5Videos 2

6See also

💬 Discuss this chapter

Ask, share, or report — over on the Heidelberg AI community forum.