Developer Documentation

Introduction

The Arshan API gives you programmatic access to everything in the platform: deploy and call models, run agents, manage datasets and receive events in real time. It is organised around REST, uses predictable resource URLs and returns JSON.

Quickstart

Install an official SDK, set your API key as an environment variable and make your first request.

# Python
pip install arshan

# Node.js
npm install @arshan/sdk

from arshan import Arshan

client = Arshan()  # reads ARSHAN_API_KEY from the environment

result = client.predictions.create(
    model="sentiment-v3",
    input="The new dashboard saves our team hours every week.",
)

print(result.label, result.confidence)  # positive 0.97

Authentication

Authenticate every request with a secret API key in the Authorization header. Create and rotate keys in Settings → API Keys. Keys are scoped to a single workspace and can be limited to read-only access.

curl https://api.arshan.ai/v1/models \
  -H "Authorization: Bearer $ARSHAN_API_KEY"

Models

List the models available to your workspace, including your fine-tuned versions.

Predictions

Send input to a deployed model and receive a prediction. Requests are synchronous for most models; long-running jobs return a job_id you can poll or receive by webhook.

Agents

Agents combine a model, tools and policies to complete multi-step tasks. Start a run, then stream steps or wait for the result.

const run = await arshan.agents.run({
  agent: "support-triage",
  input: { ticketId: "TCK-48213" },
  maxSteps: 8,
});

console.log(run.status, run.output.summary);

Webhooks

Subscribe to events such as agent.run.completed or model.deployed. Each payload is signed with your endpoint secret in the Arshan-Signature header.

{
  "id": "evt_9f2c1a",
  "type": "agent.run.completed",
  "created": 1767225600,
  "data": { "run_id": "run_51b0", "status": "succeeded" }
}

SDKs

• Python: pip install arshan (3.9+)
• Node.js: npm install @arshan/sdk (18+)
• Go: go get github.com/arshan-ai/arshan-go

Examples

Browse complete sample apps: an inbox triage agent, a RAG assistant over Google Drive, a demand forecasting notebook and a Slack bot. Each example includes a one-click deploy button.

Rate Limits

Limits depend on your plan. Responses include X-RateLimit-Remaining headers, and requests over the limit return 429. See pricing for plan limits or the changelog for recent API changes.