Platform Capabilities — Visual Builder, Multi-tenant, Billing & White-label | AI Agent Builder
Platform

What AI Agent Builder Includes

A platform to design agents, connect them to knowledge and real actions, and operate them across every customer channel. Source code included — you own the platform.

From design to production
01Create
02Connect
03Test
04Operate

Four product layers

Each layer has a distinct job and connects to the next: agent design, reliable context, actions on systems, and multichannel delivery.

01
Agent creation

Multiple ways to reach the same result: a configured, testable agent ready to operate.

  • Visual builder with workflow canvas
  • AI-assisted creation from the panel
  • Templates for repeatable cases
  • Reuse and adaptation of existing agents
02
Knowledge bases

RAG-ready knowledge so every agent can answer with real context.

  • Tenant-scoped documents and collections
  • Semantic search before answering
  • External sources like S3 and Confluence
  • Mapping per agent or use case
03
Connections

Tools that let the agent read, write and execute external processes.

  • OpenAPI for existing REST APIs
  • MCP for tools and external agents
  • Webhooks for endpoints and automations
04
Channels and widgets

The same agent is published to every touchpoint with a consistent experience.

  • Embeddable web chat
  • WhatsApp, voice and team channels
  • Widget templates for rich results
  • REST API and OpenAI-compatible endpoint for programmatic access

Action connections

The tool router can combine different sources in the same agent. OpenAPI covers existing APIs, MCP covers tools with encapsulated logic, and webhooks cover operational automations.

Operational example

A support agent can combine a product KB, a ticketing API, and a webhook to notify the team when a case needs human intervention.

OpenAPI

Use an OAS 3.x spec so the agent can discover available endpoints, parameters, and operations.

MCP

Connect tool servers or external agents compatible with Model Context Protocol.

Webhooks

Register an HTTP endpoint as a tool with body, query, path, headers, and authentication.

Multichannel delivery

Deploy the same agent across every customer touchpoint. One configuration, consistent experience everywhere.

Web Chat
WhatsApp
Voice
Slack
Microsoft Teams
Instagram
REST API
OpenAI-compatible
Integration example
<script src="https://your-domain/widget/std-chat-widget.iife.js" defer></script>
 
<script>
window.StdChatWidget.init({
apiUrl: "https://your-domain/chat",
scope: { tenantId: "my-org", agentKey: "support-agent" },
panel: { width: 420, mode: "drawer" }
});
</script>

Agent Engine

The runtime that powers your agents. Six things below are the reason teams pick this over a workflow canvas — the rest is what any serious platform should already give you.

What sets it apart
A compiled agent harness

Describe the agent and the platform builds the loop: plan the work, execute one step at a time, verify, then answer. You do not design the reasoning loop — you get one that already handles long tasks.

Loops with real exit conditions

Any group of steps can repeat until a condition you write is met, with per-iteration results kept for later steps. Bounded by an iteration limit, and it stops on its own when it detects no progress.

Automation with zero tokens

Code, direct tool calls, structured responses and cron triggers run without an LLM in the path. Same visual editor, same observability, no AI cost per execution.

Durable human-in-the-loop

Approvals, custom buttons, free text and checklists — routed to the end user or to a supervisor's inbox. The agent can also ask a question mid-task. State survives refreshes and restarts.

Headless by design

Every admin operation is available over REST and MCP. Provision tenants, deploy agents and sync knowledge from CI, or manage the platform conversationally from Claude or ChatGPT.

White-label end to end

Your brand on the widget and on the admin panel, per tenant, changed live from the UI. Source code included — your clients never see our name, and you own the platform.

Also included
Configurable agent pipeline with modular steps
Real-time streaming responses
Tool integration via OpenAPI, MCP, and webhooks
Built-in guardrails: content moderation, PII, prompt injection, topic scope
Knowledge base with semantic search (RAG)
Adaptive chunking and web content quality pipeline
Session memory and persistent long-term memory
Image generation (DALL-E 3, GPT Image, Gemini Flash Image)
Document and image attachments in chat (multimodal)
LLM resilience: circuit breaker, failover, rate limiting
Native multi-tenant with complete data isolation
Self-service billing: plans, consumption limits, Stripe checkout
Per-tenant resource limits and feature flags
Agent orchestration: sub-agents and agent-as-tool via MCP
Flow control: conditional branching, intent classification, workflow variables
Automatic KB sync from external sources (S3, Confluence, Google Drive)
Declarative widget templates for rich tool result rendering

Observability and operations

Shipping an agent is the easy half. These are the surfaces that tell you what it actually did, what it cost, whether users liked the answer, and how to roll it back when a change goes wrong.

Execution timeline

Every conversation produces a step-by-step trace with durations, inputs and results — including each tool call. You can see exactly what the agent did and in what order, not just what it said.

Cost and token tracking

Consumption and estimated cost broken down per step, conversation, agent and tenant. Find the expensive operations before the invoice does.

User feedback scoring

Thumbs up and down from end users, linked back to the exact execution trace. Turns complaints into a filterable list of runs you can open and inspect.

Health diagnostics

A running check of configuration and integration problems across the tenant — broken connections, stalled syncs, missing setup — each linking straight to the screen that fixes it.

Versioning and rollback

Every published change to an agent is a version. Browse the history and reactivate any earlier one in a click, without rebuilding the configuration by hand.

Playground

Test an agent against real tools and knowledge before it reaches a customer, with the full execution trace visible while you iterate.

Headless Admin API & MCP

The entire platform is programmable. Create tenants, agents, workflows, connections, and users via REST API or MCP — no UI required. Connect your favorite AI assistant (ChatGPT, Claude, or any MCP client) and manage the system conversationally.

Full REST API for all admin operations (CRUD tenants, agents, users, KBs, connections)
MCP server for AI-native administration — manage the platform from Claude, ChatGPT, or any MCP client
API key authentication with scoped permissions
Automate provisioning: create tenants, deploy agents, configure channels programmatically
CI/CD friendly: import/export agent artifacts, sync KB sources, apply migrations headlessly
White-label onboarding: script the entire tenant setup (plan, branding, agents, connections) via API
Programmatic access
# Create a tenant via Admin API
curl -X POST https://your-domain/admin/tenants \
  -H "x-api-key: sk_your_admin_key" \
  -H "Content-Type: application/json" \
  -d '{"name": "Acme Corp", "slug": "acme", "defaultLocale": "en"}'

# Or use the MCP server from Claude, ChatGPT, or any MCP client
mcp: create_tenant(name: "Acme Corp", slug: "acme")

Embeddable Chat Widget

A ready-to-use chat component you can embed in any website or web app. Fully customizable from the admin panel.

Embed with a single script tag — no build step required
Isolated rendering — no CSS conflicts with your site
Real-time streaming responses
Rich response cards via customizable widget templates
Image and document upload (PDF, DOCX, images) — multimodal
Image generation rendered inline with click-to-expand
Per-session tool activation — users toggle tools per conversation
Pass context from your app to the agent automatically
Public API: open, close, and control the widget programmatically
Voice input with transcription
Multi-language support (configurable per agent)
Branding and appearance customizable per agent
White-label

Full brand customization — zero code

Every tenant gets a fully branded experience — both the chat widget AND the admin panel. Pick a preset, customize colors, fonts, and logo, and the change is live instantly. Your clients see your brand, not ours. Click a preset below to see it applied to the widget on this page.

↑ Click a preset — the widget in the corner updates in real-time

Ready to see it in action?

Try the interactive agent inside the product site or explore how the pipeline works under the hood.