Use Cases
Support agent
Build a customer support agent with chat UI, tools, memory, escalation, and production guardrails.
A support agent is one of the clearest “full ecosystem” builds in AgentsKit. You usually need:
- a chat UI
- tools for ticketing, CRM, and knowledge lookup
- memory for ongoing customer context
- approval or escalation for risky actions
- observability and auditability in production
Typical stack
npm install @agentskit/react @agentskit/adapters @agentskit/runtime @agentskit/tools @agentskit/memory @agentskit/observabilityRecommended package mix
| Layer | Package | Why it matters |
|---|---|---|
| UI | @agentskit/react | Customer-facing chat surface with headless components |
| Provider | @agentskit/adapters | Swap model vendors without rewriting the app |
| Runtime | @agentskit/runtime | Multi-step reasoning, tool calling, retries |
| Tools | @agentskit/tools | CRM, helpdesk, email, browser, MCP |
| Memory | @agentskit/memory | Persistent customer context across sessions |
| Ops | @agentskit/observability | Trace runs, inspect failures, keep audit trails |
What the architecture usually looks like
- The user starts in a React chat.
- The runtime decides whether to answer directly or use tools.
- Tools fetch customer records, open tickets, or query knowledge bases.
- Memory stores prior context, preferences, and recent resolution history.
- Observability records the run for debugging, review, and compliance.
Good defaults
- Start with a conservative support skill and clear tool descriptions.
- Keep write actions confirmation-gated.
- Persist memory by customer or conversation id.
- Add audit logging before enabling actions like refunds, account changes, or outbound messages.
Best follow-up guides
- Build your first agent
- Customer support example
- Discord bot recipe
- Persistent memory
- Confirmation-gated tool
- Audit log
When AgentsKit is especially strong here
AgentsKit works well for support agents when you want the same contracts across UI, tool calls, and operations. That makes it much easier to evolve from “chatbot” into “customer-facing agent with real actions” without replacing half the stack.