Practice: Your first conversation and picking a model

Self-check

Six quick questions. Answer each in your head (or out loud) before opening the collapsible. Active retrieval beats rereading by a wide margin.

1. Name the four screen zones of Clawless and what each one is for.

Show answer

Agent rail (left, vertical strip of avatars, one per agent), chat (middle, where the conversation happens), dock row (below the input, holds frequently used controls including the model picker), navigation rail (far left, opens panels for Tools, Skills, Memory, Channels, Cron, Logs, Usage, Settings).

2. Where exactly is the model picker, and what does clicking it show?

Show answer

It is the dropdown chip in the dock row, just below the input box, displaying the current model name (for example, Claude Opus 4.6). Clicking it opens a short, curated list of models grouped by provider, with the default selection following whichever provider key you set up first during onboarding.

3. You ask a hard question, get a careful answer from a big model, then have five small follow-ups. What is the cost-saving move, and what will the agent remember?

Show answer

Switch to a smaller, cheaper model from the model picker before sending the follow-ups. The new model picks up the full conversation history and continues where the previous one left off, so nothing is lost. Costs follow the model you are using at the moment you send each message, so the meter drops immediately for the follow-ups.

4. The model you want is not on the curated short list. How do you use it anyway?

Show answer

Type the provider-prefixed name straight into the picker. The pattern is the provider slug, a slash, and the model name, as in:

openai/gpt-5.1-mini

Clawless then uses that model even if it is not on the visible short list.

5. You click between two different agents on the left rail. What happens to (a) the conversation, (b) the model selection?

Show answer

Each agent keeps its own conversation history, so clicking between them swaps you between two parallel threads. The model picker updates to that agent’s preferred model. You do not lose your place in either thread, and the picker change is intentional (each agent has its own default), not a glitch.

6. Can you switch models mid-conversation, and what does the new model see when it takes over?

Show answer

Yes. There is no per-conversation model lock. The new model picks up the full conversation history and continues where the previous one left off, so you do not have to re-paste context. The cost meter follows the model you are using at the moment you send each message, so the savings (or extra cost) take effect immediately and only forward, never retroactively.

Try it yourself: the 20-minute hands-on

The body lesson gave you the map. This is where you walk the ground. Three real messages, one mid-conversation model switch, one off-list model. Twenty minutes start to finish.

Side effects: real API calls against whichever provider you connected during onboarding. If you signed in with ChatGPT through OAuth, the off-list step uses your ChatGPT subscription (no per-token billing). If you connected an API key, the off-list step will use a small amount of credit (the cost of three to five messages on a small model is under one US cent in 2026).

Steps:

Open Clawless. Confirm you are on the Assistant agent (the default avatar at the top of the left rail). If you are elsewhere, click Assistant.
Find the model picker. Look at the dock row, below the input box. Locate the dropdown chip showing the current model name. You should be able to point at it without searching.
Send your first message. Type one of the following (or your own real question), then press Enter:
- “Help me write a short status update to my team about a project that is on track but had two small blockers this week. Keep it three sentences. Confident tone.”
- “I have a coffee chat next week with someone whose career I want to learn from. Draft five open questions I could ask, ranging from light to substantive.”
Watch the activity indicator below your message. Read the reply as it streams in.
Send a follow-up in the same conversation. Without changing the model, ask the agent to revise the reply in some way (“make the tone a bit warmer”, “drop the second bullet”, “add a sentence about timing”). Confirm that the agent remembers the original context without you re-pasting it.
Switch to a smaller model. Open the model picker. Pick a smaller model from the short list (for example, Claude Haiku 4.5 or GPT-5.1 Nano). The picker closes and the chip now shows the new model name. Send one more follow-up in the same conversation. Watch the activity indicator. Notice that the conversation continues without missing a beat, and the new model still has access to everything you and the previous model said.
Try one off-list model. Open the model picker again. Type a provider-prefixed name directly into it. Use:
```
openai/gpt-5.1-mini
```
(Substitute another off-list combination if you would rather. The format is always provider slug, slash, model name.) Send one short message and confirm the model name in the header updates to the off-list model after the reply lands.
Open the Usage dashboard. Click the Usage icon in the navigation rail on the far left. Find your conversation in the breakdown. You should see three to four entries grouped by the models you used. Total cost is in the dollars-and-cents (or zero dollars if you ran the whole thing on a ChatGPT Codex OAuth session).

Expected outcome: you finish with one conversation that has three to five messages across at least two different models, a feel for how the picker works in your hands, and a real number in the Usage dashboard for what those messages actually cost.

If something went sideways: the most common first-day surprises are listed at the end of the lesson body. Re-read those before assuming Clawless is broken.

Flashcards

Ten cards. Review once a day for a week, then on the intervals your spaced-repetition tool suggests.

Q. What are the four screen zones of Clawless?

The agent rail (left, avatars), the chat (middle, the conversation), the dock row (below the input, controls including the model picker), and the navigation rail (far left, panels).

Q. Where is the model picker?

The dropdown chip in the dock row, just below the input box. The chip shows the current model name.

Q. What is the cost rule when you switch models mid-conversation?

Costs follow the model you are using at the moment you send each message. Switching does not retroactively change any past cost; it only changes the meter going forward.

Q. Why is the model picker's list short and not exhaustive?

It is a curated short list per provider. Most providers ship dozens of models and most of the differences do not matter for most work. The short list covers capable, fast, and cheap so you do not have to choose between thirty options on day one.

Q. How do you use a model that is not on the short list?

Type the provider-prefixed name into the picker: provider slug, slash, model name. For example:

openai/gpt-5.1-mini

Q. What happens to the conversation when you switch models mid-thread?

The new model picks up where the previous one left off. It sees the entire conversation history. There is no per-conversation model lock.

Q. What does the activity indicator tell you?

What the agent is doing in real time: thinking, reading a file, searching the web, calling a tool. If you ever wonder whether the agent is stuck, the activity indicator is your answer.

Q. What is the difference between Stop and Force Stop?

Stop is the regular button (replaces the Send button while the agent is streaming) and is usually instant. Force Stop is a red banner that appears only if Stop has not taken effect after about ten seconds. Use Force Stop as a last resort.

Q. What changes when you click a different agent on the left rail?

Two things at once: the chat swaps to that agent’s separate conversation history, and the model picker updates to that agent’s preferred model. You do not lose your place in either agent’s thread.

Q. What does the model name in the chat header show, especially after you switch models mid-conversation?

The current model you have selected. If you switch models mid-conversation, the header reflects the model you have selected, not whichever model produced the previous reply.