Appearance
Run mode
Run is how you do work now — a live, conversational session with an agent that can use its full toolkit to get things done. Build is how you set up a reusable, schedulable process for work you'll do again.
On Home, the composer opens in Run mode by default. Type what you want — a question, a task, or something to automate — pick an agent (or leave it on Auto) and a model, and press Run. The agent responds in real time, calling its connected tools as needed and streaming its thinking as it works.
INFO
Run mode uses a model, so you need to be signed in — and have either a subscription or your own API key — before sending a message. See Accounts & sign-in.
A note on naming
Run mode is what earlier versions of Pencel called "Chat." It's the same live, conversational surface — the verb is now Run to make it clear you're doing the work now, in contrast to Build.
Run vs. Build
Two verbs share one composer on Home. Choose based on whether you want the work done now or set up for later.
| Run | Build | |
|---|---|---|
| What it does | Does the work now, live | Sets up a reusable, schedulable process |
| Feels like | A conversation | A short setup wizard |
| Result | An answer, an artifact, an action taken | A saved Workflow you can run on demand or on a schedule |
| Best for | One-off questions and tasks, exploring, ad-hoc work | Recurring work you'll repeat — reports, monitoring, briefings |
| Oversight | Real-time back-and-forth | Compiled steps with approval gates and guardrails |
TIP
Use Run to figure out the right approach, then Build to make it repeatable. The two complement each other — explore live, then capture what worked as a Workflow.
To run a process you've already built, you don't use Build again — you start it from the Jobs tab or let its schedule fire.
What Run mode can do
In a Run session the agent has access to its full toolkit. Depending on the agent and your guidelines, it can:
- Search your Knowledge and workspace to answer questions
- Create and edit Artifacts — documents, spreadsheets, slides, and more
- Pull data from and take action in connected services
- Kick off a Workflow on your behalf
WARNING
Agents can modify data and interact with external services. When an action needs approval — based on your guidelines and the agent's plan-approval setting — the agent pauses and asks before proceeding. See Approval workflows.
Working in a session
- Sessions — each conversation is a separate session with its own history. You can have several open with different agents and revisit past sessions any time.
- Agent selection — pick which agent to run with, or leave it on Auto and let Pencel choose. Each agent has its own instructions, knowledge, and tool access.
- Model & attachments — choose a model, attach a file, or dictate with voice from the composer controls.
- Streaming — responses appear as the agent generates them; you don't wait for the full reply.
- Execution trace — expand any assistant message to see exactly which tools the agent called and what came back.
Quick forms with /
Type / in the composer to summon a pre-built form — no AI round-trip, the same panel every time. It's the fastest way to capture structured input, like /todo to add a to-do. See Tips & shortcuts for the full list.
What's next
- Agents — choose who you run with
- Workflows — what Build produces
- Jobs — run a process you've already built
- Connections — the tools agents use in a Run