E2B
Open-source, secure environment with real-world tools for enterprise-grade agents.
Last verified:
What is E2B?
E2B is an open-source infrastructure that provides secure, isolated cloud sandboxes for AI agents and AI applications. It gives AI agents secure computers with real-world tools, allowing them to safely execute code, process data, and run tools just like humans do. The sandboxes are powered by Firecracker microVMs, providing full isolation for running untrusted workflows.
Key features include code execution in multiple languages (Python, JavaScript, TypeScript, Ruby, C++, Java, R, Bash), cloud browsers, GitHub repository access, coding tools like linters and autocomplete, audio/video editing capabilities, file upload/download, terminal commands, and internet access. E2B sandboxes can run for up to 24 hours (Pro tier), start in less than 200ms with no cold starts, and support up to 1,100 concurrent sandboxes with extra concurrency purchases.
E2B is designed for AI engineers, developers building AI agents, enterprises needing secure agentic workflows, and teams working on deep research agents, coding agents, computer use agents, data analysis/visualization, vibe coding, and reinforcement learning. It is trusted by 88% of Fortune 100 companies and has over 3.5M monthly downloads.
The platform works with any LLM including OpenAI, Llama, Anthropic, Mistral, and custom models. E2B offers Python and JavaScript/TypeScript SDKs, CLI tools, MCP integration for 200+ tools, custom sandbox templates, BYOC (Bring Your Own Cloud) options, and on-prem/self-hosted deployments.
E2B pricing
Pricing model: Freemium
E2B offers a Free Hobby tier and paid Pro/Enterprise plans with separate usage costs. Hobby tier includes: one-time $100 usage credits, community support, up to 1-hour sandbox session length, up to 20 concurrently running sandboxes, and 10 GiB free storage. Pro tier is $150/month plus usage costs, including everything in Hobby, customizable CPU (1-8 vCPU) and RAM (512MB-8GB), up to 24-hour sandbox sessions, up to 100 concurrently running sandboxes, ability to purchase extra concurrency up to 1,100, and 20 GiB free storage. Ultimate/Enterprise is custom pricing with custom solutions, custom concurrency limits, VPC deployment, dedicated support, and custom SLAs. Usage is charged per second: 1 vCPU at $0.000014/s, 2 vCPU (default) at $0.000028/s, 4 vCPU at $0.000056/s, 6 vCPU at $0.000084/s, 8 vCPU at $0.000112/s. Memory costs $0.0000045/GiB/s for 512 MiB to 8,192 MiB.
E2B pros
- Open-source infrastructure with transparent code
- Secure Firecracker microVM isolation for untrusted code
- Works with any LLM - OpenAI, Llama, Anthropic, Mistral, custom models
- Fast startup in under 200ms with no cold starts
- Supports multiple programming languages including Python, JavaScript, TypeScript, Ruby, C++, Java, R, Bash
- Up to 24-hour sandbox sessions on Pro tier
- Up to 1,100 concurrent sandboxes with extra concurrency purchases
- Python and JavaScript/TypeScript SDKs available
- 10 GiB free storage on Hobby tier, 20 GiB on Pro
- Custom sandbox templates with customizable CPU (1-8 vCPU) and RAM (512MB-8GB)
- Full filesystem, terminal, and git access in sandboxes
- Cloud browser capabilities via CDP/Kernel
- BYOC, on-prem, and self-hosted deployment options
- MCP integration for 200+ tools
- Used by 88% of Fortune 100 companies
- 3.5M+ monthly downloads and 1B+ started sandboxes
- GitHub cookbook with production use case examples
- Pay per second of running sandbox usage
E2B cons
- $150/month jump from free Hobby to Pro tier
- 1-hour session limit on free Hobby tier
- Usage costs add up with heavy compute workloads
- Customizable CPU and RAM only available on Pro tier
- Default 2 vCPU on Hobby/Pro may limit performance
- Community support only on free Hobby tier
- Extra concurrency requires separate purchase beyond 100 sandboxes
- Learning curve for custom template creation
Frequently asked questions about E2B
What is E2B and what does it do?
E2B is an open-source infrastructure that allows you to run AI-generated code in secure isolated sandboxes in the cloud. It provides AI agents with secure computers that have real-world tools like cloud browsers, GitHub repositories, coding tools, and the ability to run LLM-generated code. You control sandboxes using Python or JavaScript SDKs with methods like Sandbox.create() and sandbox.runCode().
Which programming languages does E2B support?
E2B Code Interpreter supports Python, JavaScript, TypeScript, Ruby, C++, Java, R, and Bash. If you can run it on a Linux box, you can run it in the E2B sandbox. The JavaScript/TypeScript kernel supports top-level await, ESM-style imports, and automatic promise resolution.
How long can a sandbox run?
On the free Hobby tier, sandbox sessions are limited to up to 1 hour. On the Pro tier ($150/month), sandboxes can run for up to 24 hours. Sandboxes can run for a few seconds or several hours depending on your needs and tier.
How many concurrent sandboxes can I run?
The Hobby tier allows up to 20 concurrently running sandboxes. The Pro tier allows up to 100 concurrently running sandboxes. You can purchase extra concurrency up to 1,100 sandboxes on the Pro tier. Enterprise customers get custom concurrency limits.
How is E2B priced?
E2B charges per second of running sandbox usage. The Hobby tier is free with a one-time $100 credit. Pro tier is $150/month plus usage costs. Usage pricing for Hobby/Pro: 1 vCPU at $0.000014/s, 2 vCPU at $0.000028/s, 4 vCPU at $0.000056/s, 6 vCPU at $0.000084/s, 8 vCPU at $0.000112/s. Memory costs $0.0000045/GiB/s. Storage is free up to 10 GiB (Hobby) or 20 GiB (Pro). Enterprise has custom pricing.
Can I use E2B with my own LLM?
Yes, E2B is LLM-agnostic and works with any model including OpenAI, Llama, Anthropic, Mistral, and your own custom models. The platform is compatible with any LLM and any AI agent or app. Examples in the cookbook show integration with GPT-4o, Claude, Llama 3, Codestral, Ollama, LangChain, and Llama Index.
How secure are E2B sandboxes?
Each sandbox is powered by Firecracker, a microVM made to run untrusted workflows with full isolation. Sandboxes are secure, isolated cloud environments that allow AI agents to safely execute code and process data. The infrastructure provides enterprise-grade security trusted by 88% of Fortune 100 companies.
Can I deploy E2B in my own cloud?
Yes, E2B supports BYOC (Bring Your Own Cloud), on-prem, and self-hosted deployments. It works in your AWS, GCP, or Azure account, or your VPC. Enterprise customers can get custom solutions with VPC deployment and custom SLAs.
How fast do sandboxes start?
E2B Sandboxes in the same region as the client start in less than 200ms with no cold starts. On secure quick start, sandboxes in the same region start in 80ms. This fast startup enables responsive AI agent workflows.
What support options are available?
The Hobby tier includes community support only. The Pro tier includes priority support. Enterprise customers get dedicated support. You can also join the Discord community, check the docs at e2b.dev/docs, or email [email protected] for help. Issues are typically resolved in minutes according to customer testimonials.