Chat

Chat is where you tell C5 what you need. You type in plain words. C5 figures out the rest.

Asking for something #

Type what you want and press enter. You do not need special words or a set format. These all work:

Summarize these three reports and tell me where they disagree.

Find every place in my project that still uses the old company name.

Draft a reply to this email. Keep it short and friendly.

C5 reads your request, makes a plan, and starts working. You will see the plan before the work begins, so you always know what is about to happen.

Bringing back something you sent #

Press ↑ in the message box to bring back the last thing you sent, the same way a terminal brings back your last command. Press it again to keep going further back, and ↓ to come forward again.

If you were part-way through writing something when you pressed ↑, it is not lost. Press ↓ past your newest message, or press Esc, and your half-written message comes back.

Each agent keeps its own list, and the list is still there tomorrow. ↑ only reaches for your history when the cursor is on the first line of the box, so it goes on moving the cursor normally while you are editing a longer message.

Words that change how the work runs #

Plain sentences are still all you need. But when you want to be precise, some words tell C5 how to do the job rather than what the job is. Put them at the beginning or the end of your message. C5 takes them out of the task description automatically, so the task still reads cleanly afterwards.

You can combine them:

Make a healthy meal plan, deep thinking P2

Draft this: move the old invoices into an archive folder. Quietly.

The question-mark button at the top of the chat lists the whole set, including some spellings not shown here. For what QMD is, see Memory search.

Commands #

Typing a command runs something straight away instead of starting a conversation about it. The terminal button at the top of the chat lists them, and /c5 help prints the same list into the conversation.

Pause is a stop, not a hold /c5 pause cancels the task's run rather than freezing it. The task goes back to To Do and starts from the beginning when you pick it up again.

With /c5 task and /c5 orch you can add:

• -autoreview — skip the approval step
• -intent — go through the Intent Wizard first, to sharpen a vague request
• -tools — let the agent use its tools
• -think — deep thinking
• -i — interactive: C5 asks you a few questions before it creates the task

/c5 orch -autoreview Write a haiku on cats

You can use -tools and -think on their own for an ordinary chat message, with no task created:

/c5 -tools -think What changed in this file last week?

Watching the work #

As your agents work, Chat shows you what they are doing. You will see:

• The steps in the plan, and which one is happening now
• Each tool the agent uses, and what it got back
• Any question the agent needs you to answer

Nothing is hidden. If you ever wonder why C5 did something, the answer is right there in the conversation.

Sharing files #

You can attach files to a message. Drag them onto the chat box or click the paperclip.

C5 can read most common file types — documents, spreadsheets, PDFs, images, code, and plain text. Once a file is attached, you can ask questions about it or ask for changes.

Talking instead of typing #

Click the small microphone at the left of the chat box, say what you want, and click it again to stop. Your words appear as text where you can read them over before you send.

A ring around the microphone moves with your voice while you talk, so you can tell it is hearing you before you speak a whole paragraph into nothing.

C5 can also read replies back to you — hover a reply and press the speaker button. The megaphone switch on the right-hand edge of the chat does it for every new reply automatically. It is off to begin with, which is deliberate: a long reply narrating itself in a shared office is not a pleasant surprise.

Both work out of the box with a built-in option that never sends anything off your computer, so there is nothing to set up first. Nothing listens until you press the microphone. See Voice if you want a paid provider, or want to turn the built-in ones off.

Stopping work #

If an agent is heading in the wrong direction, you can stop it. Click Stop and it stops taking new steps right away.

Stop is a brake, not an undo. Anything already done stays done — a file that was written is still written, an email that was sent is still sent. If a step needs reversing, ask for that as its own task.

You can then correct course and start again. Just say what was wrong.

When C5 asks first #

Some actions are risky — deleting files, spending money, sending things outside your computer. C5 stops and asks you before it does any of them.

You will see a clear message telling you exactly what the agent wants to do. You can allow it once, allow it every time, or say no.

You control where that line sits. See Permissions.

Picking up where you left off #

Chats are saved. Close C5, come back tomorrow, and your conversation is still there with all its history. Agents remember what you talked about — not just the transcript on your screen, but the context behind it. Restart your computer, update C5, switch which model you are using mid-conversation: the thread carries on where it left off.

If your computer restarts or the power goes out in the middle of a job, C5 picks the work back up when it starts again. You do not lose your place.

Long conversations #

You can keep digging into one subject for as long as you like. Once a thread gets long, C5 quietly keeps a running summary of the earlier part and holds the most recent exchanges word for word, so the agent still knows what you originally asked about forty questions ago.

This happens in the background, after a reply is sent — it never makes you wait.

You will see it when it does. A small Condensing badge appears in the chat header while it runs, and hovering it explains what is happening. It also leaves a line in the activity feed afterwards, alongside the tool calls, so you can look back and see when your conversation was condensed rather than wondering.

If you would rather it did not happen at all, turn off context compaction in Settings.

Getting better results #

A few habits make a big difference:

• Say what "done" looks like. "A one-page summary with three bullet points" beats "summarize this."
• Give it the files. Attach what you have instead of describing it.
• Say what to avoid. "Do not change anything in the test folder" saves you a cleanup.
• Start small. Try a small version first, see how it goes, then ask for the big one.

Where to go next #

• Productivity — watch longer jobs on a board.
• Schedules — have this happen on its own every week.
• Library — find the files your agents made.
• Voice — talk to C5 instead of typing.