Live Bridges

LangGraph live bridge

A LangGraph node can call into a Veridex agent and vice-versa; the bridge translates state and threads runId for unified tracing.

import { veridexNode } from '@veridex/agents-adapters/langgraph';
 
const graph = new StateGraph(MyState)
  .addNode('classify', classifyNode)
  .addNode('treasury', veridexNode({
    agent: treasuryAgent,
    mapState: (s) => s.userQuery,
    mergeResult: (s, r) => ({ ...s, treasuryResult: r.output }),
  }))
  .addEdge('classify', 'treasury');

The treasury node runs under Veridex governance — policy, approvals, evidence bundles — while the rest of the graph keeps running in LangGraph.

OpenAI Assistants bridge

import { handoffToVeridex } from '@veridex/agents-adapters/openai';
 
// In your Assistants tool handler:
case 'process_payment':
  return handoffToVeridex({ agent: treasuryAgent, input: args.userIntent });

Returns control to the Assistants run with the Veridex result; suspends if Veridex needs approval.

Mastra step bridge

import { veridexStep } from '@veridex/agents-adapters/mastra';
 
const workflow = workflow()
  .step(veridexStep({ agent: treasuryAgent, idAs: 'pay' }))
  .step(notifyStep);

Migration sequencing

The recommended order:

Start with the highest-stakes path. That's the one you most need policy, approvals, and audit on.
Bridge it. Most of your app keeps running where it lives.
Move read-only paths next — they have the least migration risk and let you build operator familiarity.
Remove the bridge once the workflow is fully on Veridex.

Identity & tracing across the bridge

Every bridged invocation carries the originating framework's runId as parentRunId in Veridex events. Trace inspectors render parent/child relationships so you can follow a transaction end-to-end across the seam.

Import & Export Overview