MeshFlowdocs

Search docs...⌘K

Docs/Integrations/TypeScript SDK

TypeScript SDK

The MeshFlow TypeScript SDK lets any TypeScript or JavaScript agent — whether it uses Vercel AI SDK, LangChain.js, raw fetch, or a custom loop — emit ATR-compatible spans to MeshFlow Cloud.

Install

bash

npm install @meshflow/sdk

Instrument

typescript

import { MeshFlowClient } from '@meshflow/sdk';

const mf = new MeshFlowClient({ apiKey: process.env.MESHFLOW_API_KEY });

const run = await mf.startRun({ workflowName: 'my-agent' });

Spans

typescript

await run.span({

agentName: 'researcher',

spanType: 'llm_call',

name: 'summarise',

inputText: prompt,

outputText: response,

modelVersion: 'claude-sonnet-4-6',

promptVersion: 'summarise-v3',

});

await run.end({ status: 'success' });

Context packs

typescript

await run.contextPack({

agentName: 'researcher',

sources: [{ type: 'rag', name: 'policy-db', tokenCount: 420 }],

inputTokens: 2048,

promptVersion: 'v3',

});

MCP Tools

OTel / OTLP

Docs/Integrations/TypeScript SDK

TypeScript SDK

The MeshFlow TypeScript SDK lets any TypeScript or JavaScript agent — whether it uses Vercel AI SDK, LangChain.js, raw fetch, or a custom loop — emit ATR-compatible spans to MeshFlow Cloud.

Install

bash

npm install @meshflow/sdk

Instrument

typescript

import { MeshFlowClient } from '@meshflow/sdk';

const mf = new MeshFlowClient({ apiKey: process.env.MESHFLOW_API_KEY });

const run = await mf.startRun({ workflowName: 'my-agent' });

Spans

typescript

await run.span({

agentName: 'researcher',

spanType: 'llm_call',

name: 'summarise',

inputText: prompt,

outputText: response,

modelVersion: 'claude-sonnet-4-6',

promptVersion: 'summarise-v3',

});

await run.end({ status: 'success' });

Context packs

typescript

await run.contextPack({

agentName: 'researcher',

sources: [{ type: 'rag', name: 'policy-db', tokenCount: 420 }],

inputTokens: 2048,

promptVersion: 'v3',

});

MCP Tools

OTel / OTLP