Adapters & Examples

You don't have to hand-roll AgentEvents. Adapter packages translate popular framework streams into the Sharely wire protocol, and the SDK repository ships runnable reference servers for five integration patterns.

Adapters

Vercel AI SDK — @sharelyai/adapter-vercel-ai

Wrap a streamText call in fromVercelAI(...) and you have a compliant handler in ~15 lines — swap in any provider the AI SDK supports:

import { streamText } from 'ai';
import { anthropic } from '@ai-sdk/anthropic';
import { fromVercelAI } from '@sharelyai/adapter-vercel-ai';
import type { Handler } from '@sharelyai/protocol';
 
export const handler: Handler = (input) =>
  fromVercelAI(
    streamText({
      model: anthropic('claude-sonnet-4-5'),
      messages: [
        ...input.history.map((m) => ({ role: m.role, content: m.content ?? '' })),
        { role: 'user', content: input.message },
      ],
      abortSignal: input.signal,
    })
  );

Temporal — @sharelyai/adapter-temporal

For agents orchestrated as Temporal workflows, fromTemporal({ client }) polls workflow signals and converts them into AgentEvents — giving you durable, resumable agent turns with the same streaming UX:

import { fromTemporal } from '@sharelyai/adapter-temporal';
 
export const handler: Handler = (input) =>
  fromTemporal({
    client: temporalClient,
    workflow: 'agentTurn',
    args: [input.message, input.context.threadId],
    signal: input.signal,
  });

Anything else

Any framework that exposes a stream can be adapted — yield AgentEvents from an async generator as the stream arrives (see Handler & Wire Protocol). Validate custom adapters with @sharelyai/conformance.

Runnable Examples (no API keys required)

Each example in examples/ (opens in a new tab) is a handler.ts + server.ts wired into createSharelyServer, plus a smoke.mjs that mocks the platform so you can run it offline:

git clone https://github.com/sharelyai/sharely-server-sdk.git
cd sharely-server-sdk
npm install
npx turbo run build
node examples/anthropic-sdk-direct/smoke.mjs

Live Demos (real workspace + LLM keys)

Full demo servers live in apps/ (opens in a new tab). Unlike the offline examples, these connect to a real Sharely workspace and LLM provider — copy the demo's .env.example to .env and fill in your keys:

The Temporal demos run a separate worker process alongside the server — see each demo's README for the worker command.

Choosing a Pattern

• Fastest path: Vercel AI SDK adapter — one provider-agnostic dependency, streaming and tool use included
• Maximum control: raw streaming or the direct Anthropic/OpenAI SDK patterns — you own the loop, buffering, and replay
• Durable / long-running turns: Temporal adapter — workflow retries and resumability with the same streaming UX
• Graph-based agents: LangGraph pattern — observe streamEvents from a compiled graph

Next Steps

• Quickstart — deploy and register your server
• Agent API Reference — thread, message, RAG, and registration contracts