The Agentic Harness for Production AI Systems

ChatBotKit gives teams a functional, configurable, and adoptable agent runtime. We maintain the hard parts centrally while your product teams ship focused solution code in Node.js, Go, or the stack they already trust.

Explore Agent SDK

Focused Solution Code

clear boundary

Your product code defines intent, tools, and adoption path.

Agentic Harness

shared runtime contract

Central runtime handles iteration, tool calls, events, and control.

Managed Platform

clear boundary

Conversations, observability, security, integrations, and scale.

solution code
for await (const event of execute({
  client,
  messages,
  tools: { deploy, inspect, notify },
})) {
  stream(event)
}
harness responsibilities
iteration loop
tool events
streaming
state
guardrails
observability

Harness

Centralize complexity, decentralize adoption

The harness is the repeatable layer between raw model calls and product-specific AI workflows.

Agent loop managed centrally

Planning, iteration, tool calling, exit states, and streamed events live in the harness instead of every consumer app.

Configurable tools and skills

Consumers expose precise, typed capabilities while the runtime coordinates how and when agents call them.

Adoptable by default

Start with one task, one tool, or one workflow. Expand without replacing the runtime contract.

Shared operational controls

Guardrails, policy, observability, retries, and auditability improve in one place for every adopted solution.

Portable execution model

Run stateless flows, stateful conversations, embedded agents, or service-side jobs with the same agentic shape.

Platform integration surface

Messaging, widgets, datasets, secrets, integrations, and reports connect through the same ChatBotKit platform.

How It Works

A harness your teams can adopt one workflow at a time

Product teams keep ownership of business logic while ChatBotKit owns the shared agentic runtime.

1

Define intent

Describe the job, the role, the model, and the boundaries your product workflow needs.

2

Expose tools

Attach typed functions, local utilities, platform integrations, or embedded skills.

3

Run the harness

Let the runtime coordinate planning, tool calls, streaming events, and completion.

4

Adopt safely

Roll the same contract into more products while central improvements keep compounding.

Ownership Model

The platform carries the weight. Solution code stays focused.

This is the point of the harness: teams can write focused, product-specific code without rebuilding agent infrastructure in every repo.

ChatBotKit harness
Consumer solution
Agent iteration and exitsManaged by the harness
Choose max iterations and consume events
Tool executionNormalize schema, call flow, errors
Provide focused handlers
Streaming UXEmit consistent progress and result events
Render or route events
State and contextSupport local and conversation-backed modes
Pass messages or conversation IDs
OperationsImprove guardrails, retries, logs, and reports centrally
Adopt improvements automatically

SDKs

Harness primitives for Node.js and Go

The same idea works across both ecosystems: write tools and business logic locally, then let the harness coordinate agent behavior.

Node Agent SDK

Use @chatbotkit/agent to run autonomous loops, custom Zod-typed tools, built-in tools, local execution, and conversation-backed execution.

agent.jsNode.js
1import { execute } from '@chatbotkit/agent'
2
3for await (const event of execute({
4client,
5messages,
6tools,
7maxIterations: 10,
8})) {
9publish(event)
10}
  • Async generator events
  • Zod tool schemas
  • Built-in plan, progress, exit, and abort tools

Go Agent SDK

Use the Go SDK agent package for backend services that need typed handlers, streaming events, explicit context control, and portable execution.

main.goGo
1events, errs := agent.CompleteWithTools(ctx, client,
2agent.CompleteWithToolsOptions{
3Model: "gpt-5.5",
4Messages: messages,
5Tools: tools,
6},
7)
  • Context-aware handlers
  • JSON Schema tool parameters
  • Manual or service-driven iteration

Small Product Core, Serious Agent Infrastructure

The harness gives teams a durable place to put complexity. Every improvement to the runtime becomes available to the next workflow, product, and customer.

0
agent runtimes

Node.js and Go harness entry points for service and product teams

0+
execution modes

stateless, stateful, embedded, and conversation-backed patterns

0+
shared concerns

streaming, tools, context, guardrails, logs, and integrations centralized

0%
focused

typical solution code when the harness owns the agentic plumbing

Adoption Paths

Built for teams that need agents inside real products

Internal coding and operations agents

Give trusted agents shell, file, deployment, and incident tooling while keeping the execution loop observable.

Customer workflow automations

Wrap product-specific APIs in precise tools and let the harness coordinate multi-step customer outcomes.

Regulated assistant workflows

Keep policy, auditability, and centralized controls consistent as teams adopt agentic AI across sensitive domains.

Customer Stories

Case studies from teams shipping AI

See how organizations use ChatBotKit to launch production AI assistants, validate new products, support customers, and create richer digital experiences.

Enterprise search

Quench

Discover Quench.ai, the enterprise search company founded by Husayn Kassai, the serial entrepreneur behind Onfido. Quench helps large organizations unify and discover their internal knowledge through natural language search. Built on ChatBotKit's Forward Deployment platform - the environment powering the "Quench Sandbox" - Quench prototypes, runs discovery, and validates AI products with real customers in days rather than quarters. Learn how this approach delivered 10x faster prototyping and won major enterprises including Yum Brands, MotorK, Podium, and numerous Fortune 500 companies, turning rapid customer iteration into a sustainable competitive advantage.

Forward deploymentFortune 500
Read case study

Healthcare charity

Debra

DEBRA UK is the leading charity for individuals with epidermolysis bullosa (EB), a rare genetic skin condition. Committed to providing lifelong care and seeking cures, DEBRA supports nearly 4,000 members across the UK. With over £22 million invested in research, DEBRA is the largest UK funder of EB studies. The organization addresses the complex information needs of patients and caregivers by offering reliable resources and support. Learn about DEBRA's innovative chatbot, providing 24/7 assistance for inquiries about EB, fundraising, and support services, ensuring accurate and compassionate communication. Explore DEBRA's mission to improve lives and advance research for those affected by EB.

Medical informationWebsite assistant
Read

Education wellbeing

Elggo

Discover Elggo, the MENA and Southeast Asia region's first AI-powered wellbeing platform for K–12 schools. Founded after the COVID-19 pandemic to close a gap in culturally relevant mental-health resources, Elggo delivers evidence-based curricula designed by regional psychologists and educators. By integrating ChatBotKit's conversational AI, embeddable widget, and multilingual support, Elggo provides students and teachers with always-on, personalized guidance on emotional literacy, decision-making, and growth mindset. Learn how a controlled trial of 12,000 students across 32 schools saw a 30% increase in student wellbeing, and how the platform scaled across seven countries while keeping content culturally responsive and data-driven.

K-12 schoolsMultilingual widget
Read

Cultural heritage

Faro

Discover FARO, the Flemish government's cultural heritage organization, which enhances access to heritage collections through its innovative ErfgoedApp. Launched in 2015, the app utilizes augmented reality, IoT, and AI to provide on-site, multilingual guidance for museums and heritage sites. In celebration of its 10th anniversary, FARO has partnered with ChatBotKit to introduce AI chatbots, transforming the app into an on-demand heritage guide. Visitors can ask questions about artworks and historic landmarks at any time, while geofencing technology provides location-aware storytelling. With plans to expand this interactive experience across more sites, FARO is committed to making heritage discovery intuitive and personalized for everyone.

Location-awareMuseum guide
Read

Customer service

Intelliway

Discover Intelliway, a Brazilian technology firm building AI-powered customer service solutions for businesses across Brazil and Latin America. Using ChatBotKit's API-first platform as their backend, Intelliway builds custom-branded interfaces on top of powerful conversational AI while retaining full control over the customer experience. Learn how native Brazilian Portuguese understanding, scalable cloud infrastructure, and advanced language models help Intelliway serve hundreds of clients across multiple industries, with one major retail client reporting a 40% increase in positive customer feedback. Explore how the platform-as-a-backend approach positions Intelliway to lead conversational AI across the Americas.

API and SDKsPortuguese support
Read

Ready to adopt the agentic harness?

Talk to our team. We will help you identify the first workflow, choose the right SDK, and make the runtime adoptable across your product.