​
TypeScript SDK

The TypeScript SDK package in this repository is @synapsex/sdk.

​
Install

npm install @synapsex/sdk
From source: 
cd apps/agents-synapsex/shared/sdk
npm install

​
Create A Client

import { SynapseXClient } from '@synapsex/sdk';

const client = new SynapseXClient({
  baseUrl: 'https://api.synapsex.ai',
  accessToken: process.env.SYNAPSEX_ACCESS_TOKEN!,
  tenantId: process.env.SYNAPSEX_TENANT_ID!,
});
The SDK sends:
HeaderValue
AuthorizationBearer <accessToken>
X-Tenant-IdTenant ID passed to the constructor

​
Auth

const session = await client.auth.login(
  'user@example.com',
  process.env.SYNAPSEX_PASSWORD!,
  'acme-corp',
);

client.setAccessToken(session.accessToken);
If you construct the client with refreshToken, the HTTP client can refresh the access token on 401 and retry once.

​
Agents

const agents = await client.agents.list({ status: 'running', limit: 20 });

const agent = await client.agents.create({
  vmId: 'vm-123',
  name: 'ops-agent',
  type: 'ops',
  model: 'gpt-4o',
  providerId: 'provider-123',
  tools: ['tail_logs', 'get_metrics'],
});

const task = await client.agents.dispatch(agent.id, 'Run diagnostics');
Some SDK agent lifecycle methods target endpoints that are implemented by the orchestrator gateway, not the simplified API Gateway stub. Use the gateway that exposes the matching /v1/agents/{id}/stop, /restart, and /dispatch routes.

​
Tasks

const tasks = await client.tasks.list({ agentId: agent.id, limit: 10 });
const detail = await client.tasks.get(task.id);

client.tasks.stream(task.id, {
  onMessage(event) {
    console.log(event);
  },
});

​
Billing, VMs, Incidents, Secrets, Events

The top-level client exposes resource clients for the main platform domains: 
await client.billing.getUsageSummary();
await client.vms.list();
await client.incidents.list();
await client.secrets.list();

client.events.connect();
client.events.subscribe({ type: 'agent.heartbeat' }, (event) => {
  console.log(event);
});

​
Intelligent MCP Routing

const plan = await client.mcpRouter.plan({
  input: 'Run QAOA for this graph',
  tenantId: 'tenant-123',
});

if (plan.requiresApproval) {
  console.log(plan.explanation);
}

const result = await client.mcpRouter.ask({
  input: 'Run QAOA for this graph',
  preferredMcp: 'qcos-mcp',
  approve: true,
  toolName: 'optimize_qaoa',
  toolInput: { graph: [[0, 1], [1, 2]], p: 2, shots: 1024 },
});
Use client.mcpRouter.route() for selection only, plan() for explainable previews, ask() for plan-or-execute behavior, and execute() when the caller has already collected structured inputs and approval.

​
Error Handling

Non-2xx responses throw ApiError. 
import { ApiError } from '@synapsex/sdk';

try {
  await client.agents.get('missing');
} catch (error) {
  if (error instanceof ApiError) {
    console.error(error.status, error.body);
  }
}