AI Agents
E2B
Secure sandboxes where AI agents run real code.
★ 14.2k starsApache-2.0
View guide →ByteDance's SandboxFusion: a secure sandbox for running and judging LLM-generated code in 20+ languages.
SandboxFusion is an open-source ai agents project by bytedance (bytedance/SandboxFusion on GitHub). It has 1,072 GitHub stars and 103 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 14 July 2026.
Decide whether you need just code execution or the full online-judge evaluation datasets, and which languages your workloads use.
Clone the repository and build the base and server Docker images, or set up the manual Conda and Poetry environment.
Adjust the server image's base reference as documented and start the sandbox container.
Send code-execution or judging requests to the local service and integrate it into your evaluation or RL pipeline.
Runs and returns results for snippets across Python, C++, C#, Go, Java, Node.js, TypeScript, Rust, and many more languages, with pytest, go test, JUnit, and Jest support and Jupyter-mode execution.
Supports CUDA and Python GPU workloads in addition to CPU execution.
Implements evaluation for datasets such as HumanEval, MultiPL-E, CodeContests, MBPP, MBXP, MHPP, CRUXEval, NaturalCodeBench, PAL-Math, and verilog-eval.
Documented two-image Docker build (base plus server) with host-network mode keeping the service local-only by default.
Runtime folders contain install scripts so you can add or trim language support, for example the Python runtime.
Make targets cover running with hot reload, the full test suite, individual test cases with debugging, and code formatting.
Get the sandbox source and its Docker scripts.
git clone https://github.com/bytedance/SandboxFusion.git
cd SandboxFusionBuild the base image, point the server Dockerfile at it, then build the server image, exactly as the README documents.
docker build -f ./scripts/Dockerfile.base -t code_sandbox:base .
sed -i '1s/.*/FROM code_sandbox:base/' ./scripts/Dockerfile.server
docker build -f ./scripts/Dockerfile.server -t code_sandbox:server .Host-network mode keeps it reachable only at 127.0.0.1:8080 on Linux.
docker run -d --rm --network host code_sandbox:serverIt is maintained by ByteDance's Seed foundation code team on GitHub.
Yes, it is Apache-2.0 licensed and free to self-host; you provide the compute it runs on.
No API key is needed; SandboxFusion is infrastructure that runs code for your own evaluation or training pipelines.
Simple subprocess execution is unsafe and single-language; SandboxFusion is a purpose-built, multi-language sandbox service with judging for standard code datasets built in.
Clone the repo, build the two Docker images as documented, and run the server container; the service binds to 127.0.0.1:8080.
Secure sandboxes where AI agents run real code.
A secure, fast sandbox runtime where AI agents can execute safely.
an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM
A low-code tool for building multi-agent LLM apps — chatbots, RAG and fine-tuning in one flow.
Enterprise-grade AI agent framework with multi-provider LLM management, secure knowledge bases and high-precision RAG, visual workflow orchestration, and mul…
Code, build and evaluate LLM agents with full MCP, Skills, ACP and A2A support.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if SandboxFusion is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.