Matt Brockman

.@ycombinator open-sourced the agent harness it uses internally, @qm__dev. It gives everyone on your team their own agent in Slack, so every agent needs a computer to work in. That computer can now be an E2B sandbox. Set "sandbox.backend": "e2b" and every person and channel gets an isolated sandbox from the qm-sandbox template, with the GitHub CLI, the AWS CLI, Claude Code and Codex preinstalled. Ask QM to run your tests and open a PR, and it does the work in your sandbox without touching anyone else’s. Run all of QM on E2B with e2b sandbox create qm, or point your existing deployment at it. docs.e2b.dev/agents/qm#qm

Use @OpenAI's computer-use and browser-use helpers with E2B to give your agent a private desktop in the cloud that you can watch live while the model clicks and types. Use it to help you complete tedious tasks in an application or test user flows: docs.e2b.dev/agents/openai-c…

ClickUp’s Brain² runs every frontier model with full context of your company, writing and running code to build reports, decks, and dashboards across all teams. That code touches sensitive workspace data, has to start instantly mid-conversation, and has to scale to millions of users. That’s why @ClickUp built its execution layer on E2B: every sandbox is its own microVM, starts from a prebuilt template with tools ready, and scales on demand. Enterprise-ready, from SOC 2 to HIPAA to full isolation.

Introducing E2B Secrets 🔐 Give agents API access while keeping credentials outside the sandbox. E2B injects keys into configured HTTPS request headers and lets you rotate them without restarting sandboxes. Available now! e2b.dev/resources/introducin… by @GiulioMicheloni

Why make an agent choose one way of solving a problem when it can explore them all? Catch our very own @mishushakov and @ondrej_sh at @aiDotEngineer on Thursday, September 24, as they show how agents can fork live machines to explore multiple paths in parallel- each with its own copy of the machine’s state, filesystem, and running processes.

🚀 Our open-source project has a new name: E2B Runtime. Behind every E2B stack there's a runtime. Cloud, enterprise, or embedded in your product. Same code, Apache-2.0. Old URL redirects. Something new is already in the tree. More soon. github.com/e2b-dev/runtime

Security won't affect 99% of your day-to-day with a sandbox provider. But the 1% outweighs the other 99% combined. CVE-2026-31431 is the 1%. Root on every Linux since 2017, cross-tenant on shared-kernel hosts. If your sandbox provider runs in containers, you should immediately check whether they're exposed to this. E2B isn't affected, by design. Here's why: Per-sandbox kernels. Every E2B sandbox runs in its own Firecracker microVM. Own kernel, own memory, own page cache. Bugcrowd's analysis of Copy Fail puts it cleanly: the boundaries that hold are the ones that don't share a kernel. A kernel exploit inside the sandbox cannot reach the host: the attacker would need a Firecracker escape on top of the kernel bug. The module isn't compiled in. We build custom minimal kernels for agent sandboxes. These run code interpreters, drive browsers and CLIs, power RL rollouts, and host long-running agents. CONFIG_CRYPTO_USER and CONFIG_CRYPTO_USER_API_AEAD have never been enabled, cause they're not needed by AI agents. ‍Fewer, purpose-built features mean fewer places for vulnerabilities.

We have just launched the E2B deal for YC founders. 🫡 20 YCombinator startups are joining the first batch of the deal. We are proud to support companies like Rivet, Redouble, or Prox with use cases like: 🔸 LLMs safety benchmarks 🔸 Cursor-like coding assistants for accountants, biologists, and social scientists 🔸 AI that generates, tests, and deploys full-stack apps from natural language 🔸 User simulator environments for training collaborative AI agents 🔸 RL environments for evaluating dataset quality 🔸 Cybersecurity agents that run penetration tests in sandboxed containers. See the deal is live on Bookface. To check out other E2B startup programs, look for "E2B for Startups" on our web.