Build with the AI Agentics developer platform
The complete AI agent documentation for engineers: install an SDK, register tools, run autonomous agents, and trace every decision. Go from a single API call to a production multi-agent system.
- TypeScript & Python
- REST + streaming
- Updated 2026
This is the home of the AI Agentics developer docs — the fastest path from an idea to a working agent. Whether you are wiring up your first tool call or scaling an orchestrated team of agents, every concept, endpoint, and SDK method is documented here with copy-paste examples.
The AI Agentics developer platform gives you a reasoning model, a tool runtime, durable memory, and built-in observability behind one consistent API. You bring the goal and the tools; the platform handles the agent loop — calling the model, executing tool calls, retrying on failure, and recording a full trace of every step.
New to the space? Start with what is agentic AI for the concepts, or jump straight into the quickstart below to run an agent in under five minutes. When you are ready for production, the API reference and SDKs have the full surface area.
Run your first agent in a few lines
Install the SDK, point it at a model, give it a tool, and call run(). The platform manages the loop end to end.
From zero to a running agent
The SDK exposes a single Agent primitive. Define its instructions, attach tools as plain functions, and call run() with a goal. Each run streams back the model's reasoning, every tool invocation, and the final result.
Tools are just typed functions. The platform handles JSON-schema generation, the model's tool-calling protocol, argument validation, and execution — so you write business logic, not glue code.
- One install, works in Node and the edge runtime
- Typed tools with automatic schema generation
- Streaming runs with a full trace by default
1$ npm install @aiagents/sdk23import { Agent, tool } from "@aiagents/sdk";45const weather = tool({ // a typed tool6 name: "get_weather",7 run: ({ city }) => fetchWeather(city), // your logic8});910const agent = new Agent({11 model: "agentics-pro",12 instructions: "Help users plan trips.",13 tools: [weather],14});1516const run = await agent.run("Should I pack a coat for Oslo?");17console.log(run.output); // final answer + traceGet started in 5 minutes
Five steps take you from an empty project to a traced, tool-using agent running against the platform.
1 · Install the SDK
Add @aiagents/sdk (TypeScript) or pip install aiagents (Python). Both wrap the same REST and streaming API, so concepts transfer one to one.
2 · Get an API key
Create a key in the dashboard and set AIAGENTS_API_KEY in your environment. Keys are scoped per project so dev, staging, and prod stay isolated.
3 · Define an agent
Pick a model, write clear instructions, and set guardrails. Instructions are the agent's system prompt — be specific about its goal, tone, and limits.
4 · Add a tool
Register a typed function the agent can call — a database query, a REST endpoint, or a web search. The platform generates the schema and validates arguments.
5 · Run & trace
Call run() with a goal and watch the agent loop execute. Every model call, tool invocation, and token lands in a replayable trace you can debug.
Prefer Python?
Every example in the docs has a Python tab. The Python SDK mirrors the TypeScript API: Agent, tool, and agent.run() behave identically.
Keep keys server-side
API keys carry full project access. Call the platform from your backend or an edge function, never from a browser bundle. Use short-lived scoped tokens for client-facing apps.
Steps
to a running agent
SDKs
TypeScript & Python
First run
from npm install
Traceable
every step logged
Explore the developer docs
Each section goes deep on one part of building agents. Start at the top and work down, or jump straight to what you need.
Quickstart
Start hereInstall an SDK, get a key, and run a tool-using agent in a few lines. The fastest path from zero to a working loop.
Learn moreCore concepts
The agent loop, reasoning models, planning patterns like ReAct, and how an LLM becomes an autonomous agent.
Learn moreBuilding agents
Design instructions, set guardrails, handle errors, and ship reliable agents from prototype to production.
Learn moreTools & integrations
Give agents typed tools and function calls to query databases, hit REST APIs, and act on real systems.
Learn moreMemory & RAG
Add short-term context and long-term memory with vector stores and retrieval-augmented generation (RAG).
Learn moreMulti-agent
Compose orchestrators and specialist workers that delegate, review, and coordinate on complex goals.
Learn moreObservability
Trace every decision, token, and tool call. Replay runs, debug failures, and evaluate agent quality.
Learn moreAPI reference
Every REST endpoint with schemas, auth, rate limits, streaming, and error codes for the full platform.
Learn moreSDKs
Official TypeScript and Python SDKs with the same Agent, tool, and run() surface across both languages.
Learn moreGo deeper
The most useful jumping-off points once you have your first agent running.
Want to see the platform in action first? Browse platform features, check the available integrations, or fork a starter from templates and read the code alongside the docs.
Developer docs, answered
It is the developer toolkit, SDKs, and APIs for designing, running, and observing AI agents. The documentation covers everything from a five-minute quickstart to core concepts like tool calling, memory, and orchestration, plus a complete REST API reference and language SDKs for TypeScript and Python.
Ship your first agent today
Grab an API key, install the SDK, and run a tool-using agent in minutes. Free to start — no credit card required.