Pencel

Appearance

Quickstart ​

This walkthrough takes you from a fresh install to a working agent that generates a report. By the end, you will have signed in, created an agent, authored a workflow, run it, and reviewed the result in Pulse.

Fastest path

Want a working job in one click instead of building from scratch? Open Solutions and install a solution pack (like RFP Scout or Stock Watch) — it comes with a ready-made agent, workflow, and connections. This quickstart builds one by hand so you understand the pieces.

Step 1: Sign in ​

Pencel requires you to sign in before using any model — including when you bring your own API key. On first launch you will accept the Terms of Service and sign in with Google or email; the onboarding flow then walks you through picking a role, ranking your priorities, connecting a data source, and installing a starter pack.

If you skipped onboarding or signed out, open Settings → Account and sign in. To run models you also need either an active subscription (Pencel Auto/Private modes) or your own API key — see Installation and Accounts & sign-in.

Step 2: Create an agent ​

Agents are the AI workers that carry out tasks on your behalf. Each agent has a name, a role, and a set of behaviors you define.

  1. Click Agents in the sidebar (under Manage).
  2. Click New Agent.
  3. Fill in the details:
    • Name: Report Writer
    • Role: You write concise executive summaries based on available workspace data. You focus on clarity, brevity, and actionable insights.
  4. Save the agent.

INFO

The role field is where you describe what the agent is good at and how it should behave. Write it in second person, as if you are talking directly to the agent. Be specific — "You write concise executive summaries" works better than "This agent writes reports."

Step 3: Author a workflow ​

A workflow is a multi-step process that tells an agent what to do — think of it as instructions you would give a new team member. Pencel authors workflows through a guided "Your job" pane, where you describe the outcome in plain language and Pencel drafts the steps, inputs, connections, and guardrails with you.

  1. From Home, click Build (or open Workflows → New).

  2. Describe the job in plain language, for example:

    Every morning, search the workspace for recent artifacts and documents,
summarize the key findings from the past day, and create a report titled
"Daily Summary" — organized by topic, with a brief executive overview at the top.

  3. Pencel drafts a plan. Review the collapsible Inputs, Connections, and Won't-do sections, and expand any step to see what it will do.

  4. Use Test this step to safely dry-run an individual step — a step is only marked Validated after a test passes. If a step fails, click Ask Pencel to fix.

  5. When you're happy, Finalize. Your job lands on the Jobs page.

TIP

You describe workflows in plain language — no code, no special syntax. Pencel turns your description into concrete, testable steps.

Step 4: Run it and watch it work ​

  1. Open your job on the Jobs page.
  2. Click Run once now (from the job pane or the Jobs page).
  3. Watch execution live in the Jobs run detail — each step, the tools it uses, its output, and the cost.

The Jobs surface is your run/operate home: run history, live monitoring, per-job trust rollup (clean / repaired / failed), and schedules all live here.

WARNING

Some steps pause and ask for your approval before proceeding — the human-in-the-loop gate that makes sure agents do not take actions you have not reviewed. Approve to let the agent continue.

Step 5: Review the result ​

When the run completes, its output shows up in two places:

  • Pulse — a briefing card summarizing what the agent produced. From the card you can ask a follow-up, investigate, or draft a reply without leaving Pulse.
  • Workspace — the actual report artifact. Click Workspace in the sidebar, open the Daily Summary report, and review or edit it in place.

Your agent searched the workspace, found relevant information, and produced a structured summary — all without you writing a single line of code.

What just happened? ​

Here is what Pencel did behind the scenes:

  • Context assembly. Pencel gathered the agent's role, active guidelines, relevant memory, and workspace context into a prompt.
  • Step execution. The agent worked through the workflow's steps, using built-in tools to search your workspace and create artifacts.
  • Approval gates. Any steps that modify data or produce output paused for your review before proceeding.
  • Publishing to Pulse. The result was published as a card in Pulse, and the run recorded on the Jobs page.
  • Memory. The agent learned from this interaction. Next time it runs, it will carry context from this session.

What's next ​

Now that you have a working agent, explore these guides to go deeper:

  • Core Concepts — How agents, workflows, Pulse, Jobs, Knowledge, and memory fit together
  • Creating Your First Agent — A detailed guide to agent configuration and best practices
  • Building a Workflow — Author effective multi-step workflows with the guided "Your job" pane
  • Solutions — Install a whole job in one click from the marketplace