Context Files

Context files are reference documents you upload that give agents the background knowledge they need about your business.

Where context files live

Context files are part of Knowledge (/knowledge) — the unified home for documents, context, guidelines, and connected folder vaults. They are no longer a separate surface; everything below still applies.

What It Does

Agents work best when they understand your organization. Context files are where you put that understanding — your product overview, company conventions, org chart, brand guidelines, and anything else agents should know.

Unlike memory (which agents build up over time), context files are documents you provide upfront. Think of them as the onboarding packet you would give a new team member.

Examples

Good candidates for context files include:

• Product overview — what your company does, key products, and target customers
• Company conventions — naming standards, formatting rules, approval processes
• Org chart — who is on which team, reporting lines, key contacts
• Brand guidelines — voice, tone, terminology, and style rules
• Technical specs — API documentation, system architecture, data dictionaries
• Process documents — standard operating procedures, escalation paths

Key Properties

Property	Description
Filename	Name of the context file (e.g., product.md)
Content	The text content of the file
Activation Mode	When the file is included in agent context
Source URL	Optional — fetch content from a URL
Sync Schedule	manual, daily, or weekly

Activation Modes

You control when each context file is included in an agent's prompt.

Mode	Behavior	Best For
Always	Included in every agent run	Critical context — company name, key policies, core terminology
When relevant	Included only when semantically similar to the current task	Domain-specific knowledge — product specs, department procedures
Never	Stored but not injected into agent context	Archived or draft documents you want to keep but not use yet

TIP

Keep your "always" files small and focused. Every always-on file uses part of the agent's context window, leaving less room for the actual task. Put detailed reference material in "when relevant" files instead.

URL Sync

You can link a context file to a URL. Pencel will periodically fetch the latest content, keeping your agents up to date with changing information.

• Manual — you trigger the sync yourself from Knowledge
• Daily — Pencel fetches the URL once per day
• Weekly — Pencel fetches the URL once per week

This is useful for living documents like wikis, dashboards, or shared docs that change frequently. Instead of manually updating the file, point it at the source and let Pencel handle the rest.

INFO

URL sync fetches plain text content. If your source is a web page, Pencel extracts the readable text and discards navigation, ads, and other non-content elements.

Context Files vs. Memory

These two features complement each other:

	Context Files	Memory
Source	You provide them	Agents generate them
Updates	Manual or URL sync	Automatic after runs
Content	Static reference material	Dynamic learnings and patterns
Best for	Background knowledge	Operational experience

Use context files for what your agents need to know on day one. Use memory for what they learn on the job.

What's Next

• Memory — how agents build knowledge over time
• Agents — agents use context files as part of their system prompt
• Guidelines — set rules for how agents interpret context