Ix provides a high‑performance sandbox infrastructure that can fork any AI agent environment—including full CPU, memory, storage, and network state—in about 26 ms. Each fork is a lightweight branch that can be mutated, committed, and diffed via multi‑language SDKs, while automatic memory reclamation and storage deduplication keep idle costs near zero. The platform autoscales on enterprise‑grade hardware up to 3 TB RAM, 384 vCPUs, and 1 PB storage, enabling AI research and development teams to test and iterate rapidly and cost‑effectively.
Funding
Funding not disclosed
Founders
Product
Problem
Developers of AI agents often need to test code in isolated environments that replicate full system state—including databases, files, and running processes—but provisioning such sandboxes is slow, resource‑intensive, and costly to maintain at scale.
Solution
Ix offers a sandbox infrastructure that can fork any environment in approximately 26 ms, preserving the complete state of CPU, memory, storage, and network. Each fork is represented as a lightweight branch that can be mutated, committed, and diffed against any prior state via language‑specific SDKs. The platform automatically reclaims memory from idle branches and deduplicates storage across forks, keeping idle costs near zero while supporting autoscaling up to 3 TB RAM, 384 vCPUs, and 1 PB of storage. High‑performance hardware—including EPYC Zen 5 CPUs, DDR5 ECC memory, PCIe 5.0 NVMe drives (12 GB/s), and 25 Gbps networking—ensures that AI workloads run at native speed within the sandbox.
Target Audience
Primary customers are AI research teams, machine‑learning platform engineers, and developers building autonomous agents who require rapid, reproducible, and cost‑effective sandbox environments for testing and iteration.
Features
- Sub‑millisecond environment forking that captures full system state (processes, files, databases)
- Branch‑based versioning where every mutation creates a commit and diffs can be generated between any two states
- Multi‑language SDKs (Python, TypeScript, Rust, Swift, Shell) for programmatic sandbox control
- Automatic memory reclamation and storage deduplication to minimize idle resource costs
- Autoscaling infrastructure supporting up to 3 TB RAM, 384 vCPUs, and 1 PB storage
- Enterprise‑grade hardware: EPYC Zen 5 CPUs, DDR5 ECC memory, PCIe 5.0 NVMe (12 GB/s), 25 Gbps networking