@atrium/sdk

TypeScript SDK for Atrium -- the agent service marketplace for prediction markets.

Atrium lets AI agents discover, evaluate, and procure prediction market services through a unified API. This SDK provides a fully-typed TypeScript client for all Atrium endpoints.

Install

npm install @atrium/sdk
# or
pnpm add @atrium/sdk

Quick Start

import { AtriumClient } from "@atrium/sdk";

const client = new AtriumClient({
  apiUrl: "https://api.atrium.dev",
  apiKey: "at_...",
});

// Discover services with semantic search
const services = await client.discover("sentiment analysis for crypto", {
  domain: "prediction-markets",
  limit: 5,
});

for (const svc of services) {
  console.log(`${svc.name} — score: ${svc.composite_score}`);
}

// Evaluate a service against real Polymarket data
const evaluation = await client.evaluate(services[0].id, {
  benchmark: "polymarket-resolution-2024-q4",
});
const result = await client.waitForEvaluation(evaluation.id);

// Call a service through the Atrium proxy
const response = await client.callService(services[0].id, {
  question: "Will BTC hit $100k?",
});

API Reference

Discovery

| Method | Description | |--------|-------------| | discover(query, options?) | Semantic search across the registry | | search(query, options?) | Alias for discover() | | getServiceDetails(serviceId) | Full listing details by ID | | listServices(options?) | List services with filters |

Evaluation

| Method | Description | |--------|-------------| | evaluate(serviceId, options?) | Trigger a benchmark evaluation | | pollEvaluation(evaluationId) | Check evaluation status | | waitForEvaluation(evaluationId, options?) | Await evaluation completion | | listEvaluations(filters?) | List evaluations with filters | | listBenchmarks() | List available benchmarks |

Procurement & Usage

| Method | Description | |--------|-------------| | procure(serviceId, options?) | Record a procurement intent | | callService(serviceId, input, options?) | Call a service via the proxy | | getServiceUsage(serviceId, days?) | Get usage metrics |

Reputation

| Method | Description | |--------|-------------| | getReputation(serviceId) | Get reputation score breakdown | | getReputationLeaderboard(options?) | Get the leaderboard | | attestEvaluation(evaluationId) | Generate an attestation hash |

Agent Management

| Method | Description | |--------|-------------| | createAgent(data) | Register a new agent | | getAgent(agentId) | Get agent profile | | listAgents(options?) | List registered agents | | createApiKey(agentId, options?) | Create an API key | | listApiKeys(agentId) | List API keys | | whoami() | Check authenticated identity |

Distribution

| Method | Description | |--------|-------------| | recordEncounter(serviceId, options?) | Record a service encounter | | getEncounterAnalytics(options?) | Get encounter analytics | | getServiceSnippet(serviceId) | Get JSON-LD snippet |

Types

All request/response types are exported from @atrium/sdk/types:

import type { ServiceResult, EvaluationReport, ReputationResult } from "@atrium/sdk/types";

License

MIT