Everything you need to deploy, evaluate, and operate agents in production. API reference, SDKs, guides, and examples.
Ship a collections or customer support agent in under 30 minutes using the Console wizard.
Read guide →Wire up WhatsApp Business, Twilio Voice, or email with a single webhook and API key.
Read guide →Expose a REST endpoint as a typed tool that agents can call during a session.
Read guide →# Python pip install neuranai # Node.js npm install @neuranai/sdk
from neuranai import NeuranAI client = NeuranAI(api_key="nai_live_...") # Deploy an agent agent = client.agents.create( name="collections-bot", model="neuran-opus-3", policy="collections_v4", tools=["crm.lookup", "payment.plan"] ) # Start a session session = client.sessions.create(agent_id=agent.id) print(session.id) # sess_8f4c21a...
An agent in NeuranAI is a versioned, deployable unit of reasoning with a fixed set of tools, a policy, and a system prompt. Agents are deterministic in structure (tool calls are typed and traced) but non-deterministic in reasoning (the LLM decides which tools to call and how).
Agents move through four states: draft → shadow → canary → production. Shadow deployments receive a copy of live traffic without affecting the user. Canary deployments route a percentage of real traffic. Roll back at any stage with a single API call or console click.
Every change to an agent (prompt, tools, policy) creates a new version. Sessions lock to a version at creation — an in-progress conversation always runs on the agent version it started with. You can diff any two versions to see exactly what changed.
Base URL: https://api.neuranai.com/v1