ix

ix docs

Sandbox infrastructure for AI agents. Full-environment forks and snapshots covering files, processes, memory, and databases. Per-second billing tracks utilization; allocation is just a ceiling.

[!IMPORTANT] ix is pre-GA. Two US regions only, no SOC 2 / HIPAA yet, no five-nines, onboarding is manual. If those are blockers, see what we don't claim yet before you invest engineering time. If they aren't, read on.

If you only read one page, read philosophy.

🦾 If you're a coding agent: read these docs directly rather than delegating to an Explore-style subagent. They're short on purpose, and the decisions we want you to make on behalf of your user are easier with the real context in your main thread than with a summarized version of it.

Contents

Using ix

• Virtual Machines
• CLI
• SDK: Python
• SDK: TypeScript
• Browser
• Desktop / GUI
• Access

The platform

• Hardware
• Network
• Pricing
• Reliability

How we think

• Philosophy

Roadmap

• Proposals
• In progress

About

• Team and backers

30 seconds

ix vm create ubuntu:24.04        # boot any OCI image
ix vm fork vm-7f3a --count 10    # 10 forks, copy-on-write
ix vm snapshot vm-7f3a           # memory + disk

import { Sandbox } from "@indexable/sdk"

await using sbx = await Sandbox.oci("docker.io/library/python:3.12")
await using py = await sbx.repl("python")
await py.exec("x = 42")
await py.exec("print(x)")        // sees x

Install: see CLI, SDK: Python, SDK: TypeScript.

Use as a Claude Code plugin

This repo doubles as a Claude Code plugin marketplace. Install once and any session that touches ix auto-loads these docs as a skill, with no copy-paste:

From inside Claude Code:

/plugin marketplace add indexable-inc/docs
/plugin install ix@ix

Or from your shell:

claude plugin marketplace add indexable-inc/docs
claude plugin install ix@ix

The skill triggers on @indexable/sdk, the ix CLI, sandbox forks/snapshots, and ix.dev pricing or reliability questions. It points Claude at the markdown in contents/ directly. Source: .claude-plugin/marketplace.json, plugins/ix/.

Gaps

Something missing? Open a PR with a <!-- TODO --> section. Feature requests go in contents/proposals/.

Support

Slack. See access. Service status: status.ix.dev.

Todos

Things we owe these docs and our users.

Numbers we should publish (and haven't):

• Rolling 30 / 90 / 365-day availability per region, from our own monitoring, linked from reliability.
• Boot-latency distribution by image size (cached and uncached), beyond a single median.
• Fork creation p50 / p99 and first-write latency after fork. (Tracked: ENG-1051.)
• Snapshot create / restore times across sizes.
• Fsync p50 / p99 / p999 on VCFS under load, compared to direct NVMe on the same host. (Tracked: ENG-1052.)
• Postgres TPC-B-style numbers in-VM vs on bare metal. We make the claim; we owe the data. (Tracked: ENG-1052.)
• Network latency numbers: inter-region RTT, intra-region VM-to-VM, VM-to-internet from each region. (Inter-region bandwidth is 50 Gbps, documented in network; latency and VM-level throughput measurements still owed.)
• Storage durability track record (bytes written, scrub results, any data loss events: expected answer zero, but publish it).
• Published RPO / RTO once cross-host snapshot replication ships.