CSS Customization
Theme CopilotKit components via CSS variables and class overrides.
/** * Agent factories for the Strands TypeScript showcase backend. * * `buildShowcaseAgent` is the single shared agent that serves the vast * majority of demos (the frontend differentiates each demo via * useFrontendTool / useRenderTool / useHumanInTheLoop / useAgentContext). * It mirrors the Python sibling's `build_showcase_agent` minus A2UI. * * The tool-free specialized agents (voice, byoc-hashbrown, byoc-json-render) * are mounted on dedicated sub-paths by `server.ts`. */import { readFileSync } from "node:fs";import { dirname, join } from "node:path";import { fileURLToPath } from "node:url";import { Agent, tool } from "@strands-agents/sdk";import { z } from "zod";import { StrandsAgent } from "@ag-ui/aws-strands";import type { StrandsAgentConfig } from "@ag-ui/aws-strands";import { A2UI_OPERATIONS_KEY, createSurface, updateComponents, updateDataModel,} from "@ag-ui/a2ui-toolkit";import { createModel } from "./model-factory";import { SHOWCASE_TOOLS } from "./tools";import { buildStatePrompt, salesStateFromArgs, notesStateFromArgs, stepsStateFromArgs, documentStateFromArgs, makeSubagentStateFromResult,} from "./state";import { SYSTEM_PROMPT, VOICE_SYSTEM_PROMPT, BYOC_HASHBROWN_SYSTEM_PROMPT, BYOC_JSON_RENDER_SYSTEM_PROMPT,} from "./prompts";export async function buildShowcaseAgent(): Promise<StrandsAgent> { const config: StrandsAgentConfig = { stateContextBuilder: buildStatePrompt, toolBehaviors: { // Sales pipeline lives in shared state; emit the snapshot from args. manage_sales_todos: { skipMessagesSnapshot: true, stateFromArgs: salesStateFromArgs, }, // Shared State (Read + Write) — notes panel. set_notes: { stateFromArgs: notesStateFromArgs }, // gen-ui-agent — live progress card driven by set_steps transitions. set_steps: { stateFromArgs: stepsStateFromArgs }, // shared-state-streaming — stream the document string into state. write_document: { stateFromArgs: documentStateFromArgs }, // Sub-agents — append a delegation entry carrying the actual output. research_agent: { stateFromResult: makeSubagentStateFromResult("research_agent"), }, writing_agent: { stateFromResult: makeSubagentStateFromResult("writing_agent"), }, critique_agent: { stateFromResult: makeSubagentStateFromResult("critique_agent"), }, }, }; const strandsAgent = new Agent({ model: await createModel(), systemPrompt: SYSTEM_PROMPT, tools: SHOWCASE_TOOLS, }); return new StrandsAgent({ agent: strandsAgent, name: "strands_agent", description: "A polished CopilotKit demo assistant: chat, tools, shared state, HITL, sub-agents.", config, });}/** Tool-free agent for the voice demo (transcription + basic chat). */export async function buildVoiceAgent(): Promise<StrandsAgent> { const strandsAgent = new Agent({ model: await createModel(), systemPrompt: VOICE_SYSTEM_PROMPT, tools: [], }); return new StrandsAgent({ agent: strandsAgent, name: "voice_agent", description: "Simple assistant for the voice demo — no tools.", });}/** Tool-free hashbrown UI-kit envelope generator (declarative-hashbrown). */export async function buildByocHashbrownAgent(): Promise<StrandsAgent> { const strandsAgent = new Agent({ model: await createModel(), systemPrompt: BYOC_HASHBROWN_SYSTEM_PROMPT, tools: [], }); return new StrandsAgent({ agent: strandsAgent, name: "byoc_hashbrown", description: "Hashbrown UI-kit envelope generator for the declarative-hashbrown demo.", });}/** Tool-free json-render flat-spec generator (declarative-json-render). */export async function buildByocJsonRenderAgent(): Promise<StrandsAgent> { const strandsAgent = new Agent({ model: await createModel(), systemPrompt: BYOC_JSON_RENDER_SYSTEM_PROMPT, tools: [], }); return new StrandsAgent({ agent: strandsAgent, name: "byoc_json_render", description: "json-render flat-spec generator for the declarative-json-render demo.", });}// ---------------------------------------------------------------------------// A2UI Fixed Schema (declarative-generative-ui) — dedicated backend tool.// ---------------------------------------------------------------------------//// Unlike the dynamic A2UI demo (which relies on the adapter auto-injecting a// `generate_a2ui` tool to *generate* a surface), the fixed-schema demo wires a// single plain backend tool — `display_flight` — that returns the// `a2ui_operations` envelope (createSurface -> updateComponents ->// updateDataModel). The component tree is fixed and authored ahead of time// (./a2ui_schemas/flight_schema.json); only the *data* changes per call. The// runtime A2UIMiddleware detects the envelope in the tool result and paints.// No sub-agent, no generation, no `generate_a2ui` injection.//// The schema's component names + data paths must match the showcase frontend// catalog at src/app/demos/a2ui-fixed-schema/a2ui/{definitions,renderers,// catalog}.ts — catalog id `copilotkit://flight-fixed-catalog`. This mirrors// the canonical langgraph-python demo (src/agents/a2ui_fixed.py).const _A2UI_DIR = dirname(fileURLToPath(import.meta.url));const A2UI_FIXED_CATALOG_ID = "copilotkit://flight-fixed-catalog";const A2UI_FIXED_SURFACE_ID = "flight-fixed-schema";// Fixed, pre-authored component layout. Loaded from JSON so it can be authored// and reviewed independently of the agent code.const FLIGHT_SCHEMA: Array<Record<string, unknown>> = JSON.parse( readFileSync(join(_A2UI_DIR, "a2ui_schemas", "flight_schema.json"), "utf-8"),);const A2UI_FIXED_SYSTEM_PROMPT = "You help users find flights. When asked about a flight, call " + "`display_flight` exactly ONCE with origin, destination, airline, and " + 'price. Use short airport codes (e.g. "SFO", "JFK") for ' + 'origin/destination and a price string like "$289". The tool\'s return ' + "value is an A2UI surface descriptor — the flight card is already rendered " + "to the user; do NOT call `display_flight` again for the same trip and do " + "NOT repeat the flight details in text. After the tool returns, reply with " + "one short confirmation sentence and stop.";/** * Dedicated agent for the A2UI fixed-schema demo. Returns the envelope as a * plain OBJECT (not a JSON string): the Strands TS SDK wraps an object * tool-return in a `json` content block the adapter reads and re-stringifies * into the TOOL_CALL_RESULT the client A2UIMiddleware scans for * `a2ui_operations`. (A bare string return lands in no content block and the * result comes through empty — unlike the Python SDK, which wraps strings.) */export async function buildA2uiFixedSchemaAgent(): Promise<StrandsAgent> { const displayFlight = tool({ name: "display_flight", description: "Show a flight card for the given trip. Use short airport codes " + '(e.g. "SFO", "JFK") for origin/destination and a price string like ' + '"$289". After this tool returns, the flight card is already rendered ' + "to the user via the A2UI surface — do NOT call it again for the same " + "flight; reply with one short confirmation sentence and stop.", inputSchema: z.object({ origin: z.string().describe('Origin airport code, e.g. "SFO".'), destination: z.string().describe('Destination airport code, e.g. "JFK".'), airline: z.string().describe('Airline name, e.g. "United".'), price: z.string().describe('Price string, e.g. "$289".'), }), callback: ({ origin, destination, airline, price }) => ({ [A2UI_OPERATIONS_KEY]: [ createSurface(A2UI_FIXED_SURFACE_ID, A2UI_FIXED_CATALOG_ID), updateComponents(A2UI_FIXED_SURFACE_ID, FLIGHT_SCHEMA), updateDataModel(A2UI_FIXED_SURFACE_ID, { origin, destination, airline, price, }), ], }), }); const strandsAgent = new Agent({ // Chat Completions API: the Responses adapter buffers tool-call argument // deltas, which would defeat A2UI's progressive surface streaming. model: await createModel({ openaiApi: "chat" }), systemPrompt: A2UI_FIXED_SYSTEM_PROMPT, tools: [displayFlight], }); return new StrandsAgent({ agent: strandsAgent, name: "a2ui_fixed_schema", description: "A2UI surface from a fixed, pre-authored schema (direct backend tool)", });}// ---------------------------------------------------------------------------// A2UI Dynamic Schema (declarative-gen-ui) — adapter auto-injects generate_a2ui.// ---------------------------------------------------------------------------//// Unlike the fixed-schema demo (which wires a `display_flight` tool returning a// pre-authored envelope), the dynamic demo lets the agent *generate* the// surface layout on the fly. The Next.js route// (app/api/copilotkit-declarative-gen-ui/route.ts) sets// `a2ui: { injectA2UITool: true, defaultCatalogId: "declarative-gen-ui-catalog" }`;// the runtime forwards the flag, the Strands adapter auto-injects a// `generate_a2ui` tool and drives a secondary render planner. The// `config.a2ui` block below supplies the catalog id stamped into generated// surfaces and the composition guide that teaches the planner the page's// catalog. Mirrors the ag-ui dynamic-schema reference example.//// The compositionGuide MUST describe the catalog the page registers at// src/app/demos/declarative-gen-ui/a2ui/{definitions,renderers,catalog}.ts// (catalog id `declarative-gen-ui-catalog`): Card / StatusBadge / Metric /// InfoRow / PrimaryButton / PieChart / BarChart / DataTable, composed inside// the basic catalog's Row / Column / Text (`includeBasicCatalog: true`).//// Grounding dataset + composition rules are kept in spirit with the frontend// `sales-context.ts` (SALES_DATASET + COMPOSITION_RULES) the page registers via// `useAgentContext`. The frontend context steers the PRIMARY agent; this// compositionGuide is the channel the adapter feeds to the secondary// `render_a2ui` planner (it gets `guidelines`, not the frontend App Context),// so the planner is self-contained.const A2UI_DYNAMIC_CATALOG_ID = "declarative-gen-ui-catalog";const A2UI_DYNAMIC_SALES_DATASET = `Vantage Threads (fictional B2B apparel company) — Q2 sales data. Ground every visual in these numbers; invent only plausible details consistent with them.- Quarterly revenue: $4.2M (up 12% QoQ). New customers: 186 (up 8%). Win rate: 31% (down 2pts). Avg deal size: $22.6k (up 5%).- Revenue by region: North America $1.9M, EMEA $1.3M, APAC $720k, LATAM $280k.- Monthly revenue: Jan $1.21M, Feb $1.34M, Mar $1.65M, Apr $1.38M, May $1.42M, Jun $1.40M.- Reps (vs quota): Dana Whitfield 124%, Marcus Lee 108%, Priya Sharma 97%, Tom Okafor 88%, Elena Vasquez 71%.- At-risk: total $615k ARR across 3 accounts — Northwind Retail ($340k renewal, no contact 6 weeks; severity high), Cascadia Outfitters ($180k, champion left; severity medium), Atlas Goods ($95k, stalled legal review; severity medium).- Biggest account: Meridian Apparel Group — owner Dana Whitfield, region North America, ARR $612k, renewal Sep 30, last contact 3 days ago, health green, 4 open opportunities worth $210k.- Meridian revenue by product line: Outerwear $260k, Footwear $180k, Accessories $112k, Custom $60k.`;const A2UI_DYNAMIC_COMPOSITION_RULES = `Use ONLY these exact component names (the registered catalog — any other name fails to render): Card, Column, Row, Text, Metric, PieChart, BarChart, DataTable, StatusBadge, InfoRow, PrimaryButton. The single-value KPI tile component is named exactly "Metric" (NOT "MetricTile" or "MetricCard").Pick A2UI components by the shape of the question — never ask which chart the user wants:1. Overall snapshot / "sales dashboard" → a Column (gap 16) whose first child is a Row (gap 16) of 4 Metric components (each with trend + trendValue), followed by a Row with a PieChart (revenue by region) next to a BarChart (monthly revenue, all six months Jan-Jun). Do NOT wrap the dashboard in a surrounding Card — the charts carry their own card chrome. Do NOT use StatusBadge, DataTable, or InfoRow here.2. Rep / team performance → a Column (gap 16) with a Card containing a DataTable (columns: rep, attainment, pipeline) next to or above a BarChart of quota attainment % per rep — no StatusBadge or InfoRow.3. Risk / health checks → a Column (gap 16): first a Row (gap 16) of 3 Metric components (ARR at risk $615k trend down, accounts at risk 3, biggest exposure Northwind $340k), then a Row (gap 16) with one compact Card per at-risk account (title = account name, subtitle = ARR at stake) containing a StatusBadge (error for high severity, warning otherwise) above a one-line Text with the reason and the recommended next action — no DataTable or InfoRow.4. Single account/entity details → a Row (gap 16) with a Card of InfoRow facts (owner, region, ARR, renewal date, last contact) next to a PieChart of that account's revenue by product line — no DataTable or StatusBadge.5. Part-of-whole follow-ups → PieChart; trends or comparisons over time/categories → BarChart.Compose generously — a dashboard should feel like a real analytics product, not a single widget.`;const A2UI_DYNAMIC_COMPOSITION_GUIDE = `${A2UI_DYNAMIC_SALES_DATASET}\n\n${A2UI_DYNAMIC_COMPOSITION_RULES}`;// Mirrors the langgraph-python demo's a2ui_dynamic.py SYSTEM_PROMPT.const A2UI_DYNAMIC_SYSTEM_PROMPT = "You are the embedded sales analyst for Vantage Threads, the fictional " + "B2B apparel company described in your App Context. Answer every " + "business question by calling `generate_a2ui` to draw a rich visual " + "surface, and keep the chat reply to one short sentence.\n\n" + "Ground every number in the sales dataset from App Context — never " + "invent figures that contradict it. Follow the dashboard composition " + "rules from App Context when choosing components: pick the component " + "by the shape of the question (snapshot → composed KPI dashboard with " + "charts; team performance → table; risk → status badges; single " + "account → info rows; part-of-whole → pie; trend/comparison → bar). " + "Never ask the user which chart they want. `generate_a2ui` takes no " + "arguments and handles the rendering automatically. Compose " + "generously — a dashboard should feel like a real analytics product, " + "not a single widget.";/** * Dedicated agent for the A2UI dynamic-schema demo. Wires NO `generate_a2ui` * tool — the runtime's `injectA2UITool: true` makes the adapter auto-inject it * and drive a secondary render planner to GENERATE the surface. */export async function buildA2uiDynamicAgent(): Promise<StrandsAgent> { const strandsAgent = new Agent({ // Chat Completions API: the Responses adapter buffers tool-call argument // deltas, which would defeat A2UI's progressive surface streaming. model: await createModel({ openaiApi: "chat" }), systemPrompt: A2UI_DYNAMIC_SYSTEM_PROMPT, }); const config: StrandsAgentConfig = { a2ui: { defaultCatalogId: A2UI_DYNAMIC_CATALOG_ID, guidelines: { compositionGuide: A2UI_DYNAMIC_COMPOSITION_GUIDE }, }, }; return new StrandsAgent({ agent: strandsAgent, name: "a2ui_dynamic_schema", description: "Dynamic A2UI surfaces generated on the fly (auto-injected tool)", config, });}// ---------------------------------------------------------------------------// A2UI Error Recovery (a2ui-recovery) — adapter auto-injects + runs recovery.// ---------------------------------------------------------------------------//// Same auto-injected dynamic-schema setup as buildA2uiDynamicAgent, but the// aimock fixtures force the inner render_a2ui to emit free-form/sloppy args// (heal pill) or a structurally-invalid surface on every attempt (exhaust// pill). The Strands adapter runs the toolkit validate->retry recovery loop on// its auto-inject path (default 3 attempts) and returns the// a2ui_recovery_exhausted hard-fail envelope when the cap is hit — so this// agent wires NO tool, unlike the langgraph/ADK siblings (which own the tool// explicitly via getA2UITools + injectA2UITool:false). Mirrors the ag-ui dojo// aws-strands recovery example./** * Dedicated agent for the A2UI error-recovery demo. Wires NO `generate_a2ui` * tool — the runtime's `injectA2UITool: true` makes the adapter auto-inject it, * drive the secondary render planner, and run the recovery loop. */export async function buildA2uiRecoveryAgent(): Promise<StrandsAgent> { const strandsAgent = new Agent({ // Chat Completions API: the Responses adapter buffers tool-call argument // deltas, which would defeat A2UI's progressive surface streaming. model: await createModel({ openaiApi: "chat" }), systemPrompt: A2UI_DYNAMIC_SYSTEM_PROMPT, }); const config: StrandsAgentConfig = { a2ui: { defaultCatalogId: A2UI_DYNAMIC_CATALOG_ID, guidelines: { compositionGuide: A2UI_DYNAMIC_COMPOSITION_GUIDE }, }, }; return new StrandsAgent({ agent: strandsAgent, name: "a2ui_recovery", description: "Dynamic A2UI with automatic error recovery (auto-injected tool)", config, });}What is this?#
CopilotKit has a variety of ways to customize the colors and structure of the Copilot UI components via plain CSS. You can:
- Override CopilotKit CSS variables to re-tint the whole UI
- Target the built-in class names (
.copilotKit...) for structural tweaks - Swap fonts per surface (messages, input, bubbles)
- Replace icons and labels via component props
If you need to change behavior, not just look, see slots or fully headless UI.
Scoping the theme#
The demo keeps all of its styling in a sibling theme.css file and applies it
only to the wrapper div holding <CopilotChat>. Importing the stylesheet from
the page module is enough; Next.js bundles it with the route:
import "./theme.css";Scoping every selector under a wrapper class keeps the overrides from leaking into the rest of the app.
CSS Variables (Easiest)#
The easiest way to change the colors used in the Copilot UI components is to override CopilotKit CSS variables. The demo sets them on the scope wrapper so they cascade into every nested chat component:
/* HALCYON palette — a private library at golden hour. The whole theme is * one warm parchment hue, one warm ink, and a deep copper ember used * sparingly so it actually reads as a signal. */.chat-css-demo-scope { --halcyon-paper: #f4efe6; --halcyon-paper-soft: #ece6d9; --halcyon-paper-elevated: #fbf8f2; --halcyon-card: #ffffff; --halcyon-rule: #d6cfbe; --halcyon-rule-strong: #aea48a; --halcyon-ink: #1a1714; --halcyon-ink-soft: #3d362e; --halcyon-ink-mute: #7a7468; --halcyon-ember: #c44a1f; --halcyon-ember-bright: #e45f2b; --halcyon-ember-soft: #f3d7c5; --halcyon-champagne: #98794a; --halcyon-display: "Instrument Serif", ui-serif, "Iowan Old Style", Georgia, serif; --halcyon-serif: "Fraunces", "Source Serif Pro", ui-serif, Georgia, "Times New Roman", serif; --halcyon-sans: "Inter Tight", ui-sans-serif, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; --halcyon-mono: "JetBrains Mono", ui-monospace, "SF Mono", Menlo, Consolas, monospace; --halcyon-shadow-soft: 0 1px 0 rgba(26, 23, 20, 0.04), 0 12px 32px -18px rgba(26, 23, 20, 0.18); --halcyon-shadow-ember: 0 1px 0 rgba(196, 74, 31, 0.18), 0 14px 36px -16px rgba(196, 74, 31, 0.42);}Once you've found the right variable, you can also apply the overrides inline
via the CopilotKitCSSProperties helper:
import { CopilotKitCSSProperties } from "@copilotkit/react-ui";
<div
style={
{
"--copilot-kit-primary-color": "#222222",
} as CopilotKitCSSProperties
}
>
<CopilotSidebar />
</div>Reference#
| CSS Variable | Description |
|---|---|
--copilot-kit-primary-color | Main brand/action color for buttons and interactive elements |
--copilot-kit-contrast-color | Color that contrasts with primary, used for text on primary elements |
--copilot-kit-background-color | Main page/container background color |
--copilot-kit-secondary-color | Secondary background for cards, panels, and elevated surfaces |
--copilot-kit-secondary-contrast-color | Primary text color for main content |
--copilot-kit-separator-color | Border color for dividers and containers |
--copilot-kit-muted-color | Muted color for disabled/inactive states |
--copilot-kit-shadow-sm / -md / -lg | Elevation shadows for subtle surfaces, cards, and modals |
Two token systems
The --copilot-kit-* variables above style the v1 component CSS
(@copilotkit/react-ui). The newer v2 components
(@copilotkit/react-core/v2) are Tailwind + shadcn-based and use a
separate set of design tokens. See v2 design
tokens below.
v2 Design Tokens (shadcn)#
The v2 components (@copilotkit/react-core/v2) ship a Tailwind v4 theme built
on the standard shadcn/ui token set. Instead of the --copilot-kit-*
variables, they read oklch color tokens that are scoped to
the [data-copilotkit] root and wired into Tailwind utilities through an
@theme inline block. This means you can re-skin the entire v2 UI by
overriding a handful of CSS custom properties. Every component picks the
change up automatically.
Override them on the [data-copilotkit] element (or any ancestor) the same way
you would in a shadcn project:
[data-copilotkit] {
--primary: oklch(0.55 0.22 264); /* accent / action color */
--primary-foreground: oklch(0.99 0 0); /* text on primary */
--background: oklch(1 0 0); /* surface background */
--foreground: oklch(0.145 0 0); /* primary text */
--muted: oklch(0.97 0 0); /* subtle backgrounds */
--border: oklch(0.922 0 0); /* dividers, outlines */
--radius: 0.625rem; /* global corner radius */
}
/* Dark mode is keyed off a `.dark` ancestor */
.dark [data-copilotkit] {
--background: oklch(0.145 0 0);
--foreground: oklch(0.985 0 0);
--border: oklch(0.269 0 0);
}Reference#
These are the most commonly overridden v2 tokens. Each light value has a
matching dark-mode value under .dark [data-copilotkit]. The full set
(popover, accent, destructive, chart, and sidebar variants) lives in
@copilotkit/react-core/v2/styles.css.
| Token | Description |
|---|---|
--background / --foreground | Base surface background and primary text color |
--primary / --primary-foreground | Accent/action color and the text rendered on top of it |
--secondary / --secondary-foreground | Secondary surfaces (cards, panels) and their text |
--muted / --muted-foreground | Subtle backgrounds and de-emphasized text |
--accent / --accent-foreground | Hover/active states and their text |
--border / --input / --ring | Divider/outline color, input borders, focus ring |
--destructive / --destructive-foreground | Error/danger color and its text |
--card / --popover (+ -foreground) | Elevated surface backgrounds and their text |
--sidebar-* | The sidebar's own background/foreground/border/ring set |
--radius | Base corner radius; --radius-sm/md/lg/xl derive from it |
oklch values
v2 tokens use the oklch() color space, which keeps perceived lightness
consistent across hues. You can still pass hsl(), rgb(), or hex; any
valid CSS color works.
Custom CSS#
The CopilotKit CSS is structured to allow customization via CSS classes. You can target specific pieces of the UI from your own stylesheet:
.copilotKitButton {
border-radius: 0;
}
.copilotKitMessages {
padding: 2rem;
}
.copilotKitUserMessage {
background: #007AFF;
}The demo's theme.css wraps every selector under .chat-css-demo-scope so
the overrides don't leak out. Here's the user message bubble block from
that file:
/* User message — a "transmission" in JetBrains Mono on a paper card. The * outer wrapper is the right-aligning flex column; we leave it transparent * and style the inner bubble (which uses cpk:bg-muted, hence we also * target the substring class as a stable hook). */.chat-css-demo-scope .copilotKitMessage.copilotKitUserMessage { background: transparent; padding: 0; border: none; box-shadow: none;}.chat-css-demo-scope .copilotKitMessage.copilotKitUserMessage > [class*="bg-muted"] { font-family: var(--halcyon-mono); font-size: 0.875rem; font-weight: 400; color: var(--halcyon-ink); background: var(--halcyon-paper-elevated); border: 1px solid var(--halcyon-rule); border-left: 2px solid var(--halcyon-ember); border-radius: 0; padding: 12px 16px 12px 18px; letter-spacing: -0.005em; line-height: 1.55; box-shadow: 0 1px 0 rgba(26, 23, 20, 0.03); position: relative;}/* A mono "→" marker before the user's text to read like a CLI prompt. */.chat-css-demo-scope .copilotKitMessage.copilotKitUserMessage > [class*="bg-muted"]::before { content: "→"; display: inline-block; margin-right: 10px; color: var(--halcyon-ember); font-weight: 500;}Reference#
| CSS Class | Description |
|---|---|
.copilotKitMessages | Main container for all chat messages |
.copilotKitMessage | Base class applied to every message bubble (user and assistant) |
.copilotKitInput | Text input container with typing area and send button |
.copilotKitUserMessage | Styling for user messages |
.copilotKitAssistantMessage | Styling for AI responses |
.copilotKitHeader | Top bar of chat window containing title and controls |
.copilotKitButton | Primary chat toggle button |
.copilotKitWindow | Root container defining overall chat window dimensions |
.copilotKitMarkdown | Styles for rendered markdown content |
.copilotKitCodeBlock | Code snippet container with syntax highlighting |
.copilotKitSidebar | Styles for sidebar chat mode |
.copilotKitPopup | Styles for popup chat mode |
Custom Fonts#
You can customize the fonts by updating the fontFamily property on the
relevant CopilotKit classes:
.copilotKitMessages {
font-family: "Arial, sans-serif";
}
.copilotKitInput {
font-family: "Arial, sans-serif";
}Custom Icons#
Customize icons by passing the icons prop to CopilotSidebar, CopilotPopup,
or CopilotChat:
<CopilotChat
icons={{
openIcon: <YourOpenIconComponent />,
closeIcon: <YourCloseIconComponent />,
}}
/>Custom Labels#
Customize all user-facing copy via the labels prop:
<CopilotChat
labels={{
welcomeMessageText: "Hello! How can I help you today?",
modalHeaderTitle: "My Copilot",
chatInputPlaceholder: "Ask me anything!",
}}
/>