How It All Fits Together

Pencel is built around a small set of concepts that work together to automate your business operations.

Before diving into each concept individually, it helps to see the big picture. The diagram below shows how every piece connects.

The Concept Map

Workflow ──── assigned to ────→ Agent
Agent ──── governed by ────→ Guidelines (in Knowledge)
Agent ──── uses ────→ Connections (in Settings)
Agent ──── draws on ────→ Knowledge (docs, context, memory, vaults)
Agent execution ──── produces ────→ Artifacts (Workspace)
Agent execution ──── runs on ────→ Jobs (history, monitoring, schedules)
Triggers ──── schedule or start ────→ Jobs
Results ──── published as cards to ────→ Pulse (Feed + To-dos + Radar)
Solution packs ──── install ────→ Agents + Workflows + Connections at once

Read it top to bottom: you author agents and workflows (or install a solution pack from the Marketplace), they run as Jobs, and their results land in Pulse for you to read and act on.

What Each Relationship Means

Workflow → Agent Every workflow is assigned to an agent that carries out the work. You pick the agent when you create the workflow, though you can change it later. Think of the workflow as the script and the agent as the actor.

Agent → Guidelines Guidelines shape how an agent behaves. They set the tone, enforce data policies, and draw boundaries around what the agent can decide on its own. A single guideline can apply to every agent in your workspace or just one specific workflow.

Agent → Connections Connections are the external tools and services your agent can reach — Slack, Notion, Google Sheets, a database, and more. Without a connection, an agent can only think and write. With connections, it can actually do things.

Agent → Memory Memory lets an agent retain lessons learned across runs. If an agent discovers that a particular client prefers bullet-point summaries, it remembers that for next time. Memory is automatic — you do not need to configure it.

Agent execution → Artifacts Artifacts are the tangible outputs of a run: reports, spreadsheets, drafted emails, or any other file the agent produces. They are saved and versioned so you can review or reuse them later.

Agent execution → Runs Every time a workflow executes, Pencel creates a run. Runs track each step, show you what the agent did, how long it took, and whether it succeeded. They are your audit trail.

Triggers → Runs Triggers determine when a workflow runs. You can trigger a run manually, on a schedule (e.g., every Monday at 8 AM), or in response to an event.

Knowledge → Agent Knowledge is where agents get their background: documents, context files, guidelines, memory, and connected folder vaults, organized into a knowledge graph. You give the agent domain knowledge — company policies, product catalogs, SOPs — so it can make informed decisions without you repeating yourself. (Guidelines and context files used to be separate surfaces; they now live inside Knowledge.)

Results → Pulse Every run publishes its result as a card in Pulse — a briefing card, or a roll-up card that routes into a pack's record tab. Pulse is where you read what happened and act on it (approve, ask, investigate, draft a reply).

TIP

You do not need to set up everything at once. Most teams start by creating one agent, writing a simple workflow, and running it manually. Add guidelines, connections, and schedules as you learn what you need.

New to roles and jobs?

If you are unsure when to use an agent vs. a workflow, read Agents vs Workflows: Roles and Jobs for the two-question decision funnel and worked examples.

Explore Each Concept

Surfaces — the places you work:

Surface	What it is	Learn more
Pulse	Your home for reading results and acting on them	Pulse
Jobs	Run history, live monitoring, trust rollups, schedules	Jobs
Workflows	The reusable jobs you author (a workflow is the app)	Workflows
Agents	AI workers with distinct roles and permissions	Agents
Knowledge	Docs, context, guidelines, memory, folder vaults	Knowledge
Marketplace	Marketplace of solution packs that install a whole job in one click	Marketplace
Workspace	Files and outputs produced by runs	Artifacts & Workspace

Building blocks — the concepts inside those surfaces:

Concept	What it is	Learn more
Agents vs Workflows	When to use which — the role/job mental model	Agents vs Workflows
Guided Task	The "Your job" pane you author workflows in	Guided Task
Steps	The building blocks inside structured workflows	Steps
Triggers & Schedules	When and how workflows run (Heartbeat, cron)	Triggers & Schedules
Guidelines	Policies that shape agent behavior (in Knowledge)	Guidelines
Context Files	Domain knowledge documents (in Knowledge)	Context Files
Memory	How agents retain and recall information (in Settings)	Memory
Connections	External tools and services agents can use (in Settings)	Connections
Custom Tools	Reusable JavaScript functions agents can call	Custom Tools
To-dos	The personal checklist scope in Pulse	To-dos
Radar	Opt-in background scans that post insight cards	Radar
Runs	The execution log for every workflow	Runs
Chat	Direct conversation channel with agents	Chat
Tailor	The RFP proposal builder	Tailor