Fournex
RL-as-a-Service
REST API
Multiple Frameworks

Deploy Your RL Agents in Minutes

In supervised ML, deployment means serving predictions. In reinforcement learning, it means serving interactions — tracking state, executing rollouts, and managing environments. Fournex handles the complexity so you can focus on training better agents.

Upload Any Agent

Support for .zip, .onnx, .pt, and Stable Baselines3 checkpoints.

State Management

Automatic tracking of state, action histories, and sessions.

Gym or REST Envs

Define environments using Gym API or custom REST endpoints.

Production Ready

Scalable infrastructure with monitoring and logging built-in.

Quick Start

REST API

Upload Agent

curl -X POST https://api.fournex.com/agents/upload -F 'model=@agent.zip'

Deploy Endpoint

curl -X POST https://api.fournex.com/agents/deploy -d '{"agent_id":"abc123"}'
# Upload your trained agent
curl -X POST https://api.fournex.com/agents/upload \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "model=@my_agent.zip" \
  -F "framework=stable-baselines3" \
  -F "environment=CartPole-v1"

# Response
{
  "agent_id": "agent_abc123",
  "status": "uploaded",
  "endpoint": "https://api.fournex.com/agents/agent_abc123/predict"
}

Supported Formats

4+

.zip • .onnx • .pt • Stable-Baselines3

Deployment Time

< 5 min

from upload to production endpoint

API Type

REST

JSON over HTTPS

RL Deployment
Beta
REST API • Managed Infrastructure

Why Fournex?

Reinforcement learning deployment is fundamentally different from supervised ML. We handle state tracking, environment execution, and policy rollouts so you can focus on building better agents.

The Deployment Problem

RL isn't just about predictions — it's about managing stateful interactions.

  • Track state and action histories per session
  • Execute policy rollouts against environments
  • Log rewards and states for online learning
  • Handle evolving environments and contexts

Upload Any Agent

Support for multiple frameworks and model formats out of the box.

  • Stable Baselines3 (.zip checkpoints)
  • ONNX models (.onnx)
  • PyTorch checkpoints (.pt, .pth)
  • Custom serialized agents
SB3
ONNX
PyTorch

Environment Definition

Define your environment using standard APIs or custom endpoints.

  • Gym API compatibility (built-in envs)
  • Custom REST endpoints for your environment
  • Environment state management handled
  • Support for discrete and continuous spaces

Production Endpoints

Deploy as REST APIs with automatic scaling and monitoring.

  • REST API with JSON payloads
  • Automatic session and state management
  • Built-in monitoring and logging
  • Scalable infrastructure (auto-scaling)

State & Session Tracking

Automatically manage per-client state, action histories, and context.

  • Per-session state isolation
  • Action history tracking
  • Episode management and reset handling
  • Context persistence across requests

Online Learning Ready

Log rewards and observations for continual training workflows.

  • Reward logging for all interactions
  • State transition recording
  • Export data for retraining
  • A/B testing support for multiple agents

Fournex eliminates the infrastructure complexity of deploying RL agents. Upload your trained model, define your environment, and get a production-ready REST endpoint in minutes. We handle state tracking, session management, and scalability — you focus on building better agents.