# Providing Context to Copilot

The AI Copilot within the embedded workflow builder in your app can be provided with additional context to help it generate better results as it builds workflows for your customers.

To provide context to Copilot, add an `agent.context` object to your `prismatic.showWorkflows()` or `prismatic.showWorkflow()` calls. The `agent.context` object is a `Record<string,string>` object where record keys are file names, and record values are the file file contents.

Provide a glossary and logging conventions to Copilot

```ts
prismatic.showWorkflow({
  workflowId,
  selector: `#${embeddedDivId}`,
  agent: {
    context: {
      "glossary.md": [
        "# Glossary",
        "",
        'A "ticket" is a support request from an end customer. Ticket fields',
        "arrive on the webhook trigger body as `body.ticket`.",
        "",
        'An "audit trail" is a durable record of what a workflow did. It is',
        "always a single Log component **Write Log Message** step whose message",
        "begins with the literal prefix `AUDIT:`.",
      ].join("\n"),

      "conventions/logging.md": [
        "# Logging conventions",
        "",
        "- Every audit trail entry is one Write Log Message step.",
        "- Never add a branch step just to write an audit trail.",
      ].join("\n"),
    },
  },
});

```

The files you provide in the `agent.context` object are not immediately injected into the system prompt for Copilot, but Copilot is instructed to read from them to gather information about context. Both the key (filename) and the content are helpful in directing the Copilot when to read these. If it's helpful, this context is injection is loaded by progressive disclosure (very similar to how Claude or other AI coding agents load context as you ask questions about it).