AgentENV logo

AgentENV

AgentENV (AENV) is a distributed platform for running agent environments at scale.

RustAI AgentsDeveloper ToolsRustMIT 2.4k 183Health 85
GitHub

What is it?

What it is

AgentENV (AENV) is a platform for running agent environments at scale, powering agentic RL training.

Why it exists

To enable scalable, low-cost, high-performance execution of agent environments via fast snapshotting, forking, overlaybd caching, and efficient resource reclamation.

Who should use it

AI researchersMachine Learning engineersReinforcement Learning practitionersDevOps engineers building agentic workflowsDevelopers needing high-scale sandboxed environments

Who should avoid it

Users requiring built-in authorization/security layersUsers on kernels older than 6.8Users on non-Linux operating systems (for the server component)

How it works

A quick walkthrough in plain English

How AgentENV works

Step 1 of 3

You interact with it

Open AgentENV, send a request, or connect it to your stack.

Features

Firecracker microVM execution
On-demand OCI-compatible image loading via overlaybd
Incremental memory and filesystem snapshotting
Native sandbox forking for parallel workflows
E2B-compatible HTTP API
Memory ballooning for high overcommit density
High-performance I/O via ublk
S3-compatible object storage persistence for snapshots

Advantages

  • Rapid startup and resume times (under 50ms/100ms)
  • Scalable environment management across diverse machines
  • Efficient resource utilization through shared host page cache
  • Reduced infrastructure costs by making idle environments inexpensive
  • High density through memory reclamation and bounded disk caching
  • Seamless integration with existing E2B Python/TypeScript SDKs

Disadvantages

  • Requires modern Linux kernel (6.8+)
  • Currently lacks built-in authorization/security controls
  • Requires privileged access to /dev/kvm
  • Installation script is specifically optimized for Ubuntu 24.04

Installation

docker

curl -fsSL https://raw.githubusercontent.com/kvcache-ai/AgentENV/main/scripts/docker-setup.sh | sudo bash
docker pull ghcr.io/kvcache-ai/aenv-server:latest
docker run -d --privileged -v /dev:/dev -p 8000:8000 ghcr.io/kvcache-ai/aenv-server:latest

native

curl -fsSL https://raw.githubusercontent.com/kvcache-ai/AgentENV/main/scripts/install.sh | sudo bash
sudo systemctl start aenv

compose

See 📖 [Deployment](https://kvcache-ai.github.io/AgentENV/deployment/manual-compile.html) for Docker Compose / Kubernetes cluster deployment.

FAQ

How do I install AgentENV on Ubuntu 24.04?

Use the install script provided in the repository. Run: ```bash curl -fsSL https://raw.githubusercontent.com/kvcache-ai/AgentENV/main/scripts/install.sh | sudo bash sudo systemctl start aenv ``` This installs both the server and the `aenv` CLI, then starts the server as a systemd service. The server will be available at `http://127.0.0.1:8000` by default.

Can I run AgentENV using Docker instead of the install script?

Yes. First pull the server image and run it with privileged access: ```bash curl -fsSL https://raw.githubusercontent.com/kvcache-ai/AgentENV/main/scripts/docker-setup.sh | sudo bash docker pull ghcr.io/kvcache-ai/aenv-server:latest docker run -d --privileged -v /dev:/dev -p 8000:8000 ghcr.io/kvcache-ai/aenv-server:latest ``` The server will be reachable at `http://127.0.0.1:8000`.

How do I pull a template and start a sandbox?

First pull an OCI-compatible image as a template: ```bash aenv pull ubuntu:22.04 --name ubuntu ``` Then start a sandbox from that template: ```bash aenv start ubuntu ``` This will launch a Firecracker microVM and attach an interactive shell. Use `--detach` to start without attaching.

What is the authentication flow for the `aenv` CLI?

Run `aenv auth` to configure the CLI. It will prompt for the server URL (default `http://127.0.0.1:8000`) and an API key. The key is currently a placeholder (`dummy`), but you should supply a valid key if you have one. Once authenticated, subsequent commands will use this configuration.

How can I use AgentENV with the E2B SDK?

AgentENV implements an E2B-compatible HTTP API. Set the environment variable `E2B_API_URL` to point at your AgentENV server, e.g.: ```bash export E2B_API_URL=http://127.0.0.1:8000 ``` Then use the standard E2B Python or TypeScript SDK; no code changes are required.

Loading documentation…
View on GitHub

Featured in Videos

YouTube tutorials and walkthroughs for AgentENV

Alternatives

Similar projects ranked by category, topics, and text overlap.

Compare
AgentENV | MushyBook