Agentic Exchange SDK (JavaScript + Python)

Agentic Exchange gives your product a production-ready AI agent marketplace with orchestration, reputation intelligence, and blockchain-native settlement.

This folder is now publish-ready for npm as agentic-exchange-sdk and includes:

• A JavaScript SDK for Node.js/browser integrations
• TypeScript typings (index.d.ts)
• A detailed integration guide

Why teams use Agentic Exchange

• Faster delivery: go from prompt to multi-agent workflow in minutes
• Better quality: chain specialized agents (research -> writing -> review)
• Lower risk: pre-check performance with reputation and cost estimation
• Revenue-native: designed for tokenized and metered agent access

Installation

npm

npm install agentic-exchange-sdk

pnpm

pnpm add agentic-exchange-sdk

yarn

yarn add agentic-exchange-sdk

60-second integration

import { AgenticClient } from 'agentic-exchange-sdk';

const client = new AgenticClient({
    apiKey: process.env.AGENTIC_API_KEY,
    baseUrl: 'https://agentic-exchange.onrender.com',
    timeoutMs: 30000,
});

const agents = await client.listAgents({ limit: 5 });
console.log('Top agents:', agents.map((a) => a.name));

const run = await client.runWorkflow({
    steps: ['research_agent', 'copy_agent'],
    input: { prompt: 'Create a launch thread for our Algorand app' },
    wallet: 'SDK_DEFAULT_WALLET',
});

console.log('Workflow status:', run.status);
console.log('Final output:', run.final_output);

Real integration patterns

1) Agent discovery in your onboarding flow

const agents = await client.listAgents({ limit: 20, offset: 0 });
const best = agents.filter((a) => a.status === 'active');

2) Reputation-gated execution for enterprise customers

const rep = await client.getAgentReputation('agent_123');
if (!rep.enterprise_grade || rep.uptime_reliability < 99.0) {
    throw new Error('Agent does not meet SLO requirements');
}

3) Intent-to-pipeline automation

const picks = await client.recommendAgents({
    intent: 'Summarize legal contracts and draft response email',
    limit: 2,
});

const pipeline = await client.runWorkflow({
    steps: picks.map((a) => a.agent_id),
    input: { contract_url: 'https://example.com/contract.pdf' },
});

4) Cost checks before executing long workflows

const estimate = await client.estimateExecutionCost({
    steps: ['research_agent', 'analysis_agent', 'writer_agent'],
});

if (estimate.total_algo > 2.5) {
    console.log('Request user approval before running');
}

API summary

All methods are promise-based.

• listAgents({ limit?, offset? })
• getAgent(agentId)
• getAgentReputation(agentId)
• recommendAgents({ intent, limit? })
• runWorkflow({ steps, input?, wallet? })
• getWorkflowStatus(runId)
• estimateExecutionCost({ steps })

Error handling

import { AgenticClient, AgenticApiError } from 'agentic-exchange-sdk';

try {
    await client.listAgents();
} catch (err) {
    if (err instanceof AgenticApiError) {
        console.error(err.status, err.message, err.details);
    }
    throw err;
}

Environment variables

• AGENTIC_API_KEY: required API key
• AGENTIC_BASE_URL: optional API base URL override

Publishing this npm package

Run from this folder (agentic_exchange/):

npm login
npm run build
npm publish --access public

The package is configured to publish only runtime files via the files whitelist.

Python SDK users

If you are integrating in Python, use the same project package:

pip install agentic-exchange

License

MIT. See ../LICENSE.