Build & ship an agent in minutes

2. Core Concepts

4. Add Tools

Tools let the LLM call out for data or actions. You provide the surface; the model decides when to use a tool and how to weave returned data into the answer.

A. OpenAPI Import

1. Click Import OpenAPI.
2. Paste a spec URL or upload JSON/YAML.
3. Select the operations you want; they become tools instantly.
4. Optionally disable noisy endpoints later.

B. Custom Tool

1. Define: name, description (what it returns / when to use), HTTP method & path.
2. Optional headers & JSON body (static or with simple placeholders).
3. Attach credentials (configured in Settings) — they’re injected server-side.
4. Save; enable it for the agent.

Return plain JSON or text. If you return a recognized display block, Ordo renders visuals automatically.

Display Blocks (Optional)

For visuals, your tool response body can directly return:

{ "type":"chart", "variant":"line", "data":{ "labels":["Jan","Feb"], "datasets":[{"label":"MRR","data":[12000,14000]}] } }

{ "type":"diagram", "format":"mermaid", "code":"graph TD; A-->B; B-->C;" }

No extra wiring required — the transcript shows the chart/diagram inline.

7. Test & Iterate

• Use the agent preview chat to exercise tool paths.
• Check Chat History → select a session → inspect tool calls & tokens.
• Refine system prompt & tool descriptions for better selection quality.
• Add prompt snippets for reusable instructions (e.g. “Return a concise executive summary”).

9. Web Embed

The production embed uses a lightweight client script plus a window._ordochat config object. Provide only non-sensitive values here (never raw model provider keys).

<!-- Ordo Chat (Production Embed) -->
<script>
  window._ordochat = {
    apiKey: 'YOUR_PUBLIC_ORDO_KEY',        // Scoped Ordo key (not your LLM key)
    instanceName: 'your-agent-identity',   // Matches the agent identity you created
    apiUrl: 'https://pub.ordo.chat/api',                // Public widget API (proxy)
    autoLoadConfig: true                   // Fetch remote display & tool config automatically
    // Optional advanced overrides:
    // theme: 'dark',                      // light | dark | blue | red | black-gold
    // position: 'bottom-right',           // bubble position
    // startMaximized: false,
    // showBubble: true,                   // show floating bubble
    // title: 'Assistant',
    // placeholder: 'Ask a question...'
  };
</script>
<script src="https://ordo.chat/chat-agent-client.js" async></script>

Key fields:

• apiKey: Public, scoped Ordo key intended for embedding; safe to include on any public site. You can revoke/regenerate it in Ordo if you ever need to disable it. Never paste raw OpenAI (or other provider) keys here.
• instanceName: The agent identity (e.g. support-bot) you configured in the Agent Builder.
• autoLoadConfig: When true, theme + personality details are fetched server-side; you can still override via optional properties.

The script injects the chat bubble and handles sessions, tool invocation, and visual block rendering. All sensitive execution (credentials, outbound calls, RAG) stays server-side.

Lifecycle Events

The widget dispatches chatagent:* events on window.

window.addEventListener('chatagent:load', (e) => {
  // e.detail = { sessionId, instanceName, headless }
});
window.addEventListener('chatagent:message-sent', (e) => {
  // e.detail = { message, sessionId, instanceName }
});
window.addEventListener('chatagent:message-received', (e) => {
  // e.detail = { message, sessionId, instanceName }
});

11. Security & Governance

• Server-side credential storage & redaction.
• SSRF guard rails: outbound host allowlist (configurable).
• Per-tool rate limits (plan-based + custom overrides).
• Audit trail: every tool call & response logged per session.
• RBAC: separate admin vs member permissions.

Enterprise features (SCIM / SSO / advanced retention) available on higher plans (see pricing).

13. Plans & Limits

Core, Pro, and Scale tiers define monthly quotas (agents, conversations, pages ingested, web lookups). Annual billing applies a discount. See the pricing section for current numbers.

Overages: upgrade or add Boost Packs (+25% each) where available.