---
title: "Agent Framework"
description: "Build multi-agent systems with AINative Agent Framework — orchestration, tools, and memory"
canonical: "https://docs.ainative.studio/docs/agent-framework/overview"
last-updated: "2026-10-22T17:48:46.689Z"
---

# Agent Framework

Source: https://docs.ainative.studio/docs/agent-framework/overview

> Build multi-agent systems with AINative Agent Framework — orchestration, tools, and memory

## Agent Framework

The AINative Agent Framework lets you build, deploy, and orchestrate autonomous AI agents that run on the platform. Agents can be composed into multi-agent swarms, equipped with plugins that extend their capabilities, and given durable state that persists across invocations.

**Base URL:** `https://api.ainative.studio`

**Authentication:** All endpoints require `Authorization: Bearer <your_api_key>` unless noted otherwise.

---

## What you can build

- **Autonomous coding agents** — Agents that analyze code quality, generate fixes, and manage their own sessions.
- **Multi-agent swarms** — Submit a task; the platform routes it to a coordinated team of specialist agents (architect, backend, QA, devops, data, security, docs, frontend).
- **Plugin-extended agents** — Install capabilities from the plugin registry onto individual agents at runtime.
- **Stateful long-running agents** — Agents with durable ZeroDB-backed storage, workspace snapshots, and configurable TTLs.

---

## Core concepts

| Concept | Description |
|---------|-------------|
| **Platform Agent** | An infrastructure-level agent registered via `POST /agents/register`. Has its own `agent_id` and `api_key`. |
| **Swarm Task** | A unit of work dispatched to a multi-agent swarm. Tasks are queued and polled asynchronously. |
| **Plugin** | A capability package from the plugin registry that can be installed on a specific agent. |
| **State** | ZeroDB-backed NoSQL tables scoped to an agent. Isolated per agent, supports key-value queries. |
| **Snapshot** | An atomic point-in-time capture of all agent state — memory, KV store, conversation history. |
| **Trust Score** | A computed 0–100 score reflecting an agent's reliability, responsiveness, and compliance history. |

---

## Sub-pages

- [Agents](./agents) — Register, manage, plugins, and state management
- [Swarms](./swarms) — Multi-agent task orchestration

---

## Quick example

Register a platform agent and immediately submit a swarm task using its ID:

```bash
## 1. Register an agent
curl -X POST https://api.ainative.studio/api/v1/public/agents/register \
  -H "Content-Type: application/json" \
  -d '{
    "agent_name": "my-coding-agent",
    "capabilities": ["code_quality", "refactoring"],
    "callback_url": "https://example.com/webhooks/agent"
  }'

## 2. Submit a swarm task using the registered agent
curl -X POST https://api.ainative.studio/api/v1/agent-swarm/tasks \
  -H "Authorization: Bearer <your_api_key>" \
  -H "Content-Type: application/json" \
  -d '{
    "description": "Refactor the auth module to use dependency injection",
    "registry_agent_id": "<agent_id_from_step_1>"
  }'
```