Beam
Persistent cloud VMs, sandboxes and GPUs for AI agents
Linked from beam.cloud · checked 29 Oct
- Tool type
- Cloud VMs, sandboxes and GPU compute
- Made for
- Developers building AI agents
- Runs on
- Beam cloud, accessed through CLI, SDK or MCP
- VMs, sandboxes and apps run on Beam's cloud; drive them via CLI, SDK, or MCP.
What it does
Beam is a cloud compute platform for AI agents and the developers who build them. You create persistent Linux virtual machines, disposable sandboxes, or serverless GPU endpoints, and drive them from Python or JavaScript SDKs, the Beam CLI, or an MCP server. Inputs are code, container images, and commands; outputs are running machines, desktops, preview URLs, and deployed inference APIs.
Typical uses include giving a coding agent a dev box with SSH and a durable disk, running untrusted agent-generated code in sandboxes that can fork from snapshots, and deploying Python models as autoscaling endpoints. Memory snapshots let paused machines resume with RAM and processes intact in about a second, and VM compute is billed per second only while the machine runs.
Beam suits developers and teams running agent workloads or GPU inference who want usage-based billing. Access starts with an install script that sets up the CLI and opens a browser sign-in. Reserved GPU clusters are arranged by contacting Beam, and the open-source Beta9 runtime does not include every Beam Cloud feature.
What it can do
- Persistent Linux VMs with durable root disk, SSH, and stable URLs
- Disposable sandboxes for running untrusted code, with GPU attachment and snapshot forking
- Memory snapshots that pause and resume VMs with RAM and processes intact
- Serverless GPU endpoints that deploy Python functions as autoscaling APIs
- Egress proxy that keeps real API keys off the VM and allowlists outbound hosts
- Linux desktop that agents can screenshot, click, and type into through the SDK or MCP
Works with
- Cursor
- Claude Code
- Codex
- GitHub Actions
Before you choose
- VMs, sandboxes and serverless compute are billed per second or millisecond; storage is free up to 1 TB, then $0.021 per GB per month.
- GPU clusters are reserved and arranged by contacting Beam.
- The open-source Beta9 runtime does not include every Beam Cloud feature.
Listed under
Get started
Run the install script, which opens a browser sign-in.
- Run the install script at https://www.beam.cloud/install to install the Beam CLI and sign in through the browser.
- Choose a quickstart in the docs, such as creating a persistent VM for a coding agent.
- Connect a coding agent like Cursor, Claude Code, or Codex so it can create and operate VMs.
Plans and prices
Prices in USDCompute is billed per second while running; stopped VMs incur no compute charge. Unused monthly credits expire at renewal. Storage is free up to 1 TB, then $0.021 per GB per month.
- $20 in credits each month
- 1 seat
- 3 seats included, $25 per additional seat
- Live chat support
Source: beam.cloud/pricing, checked 29 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 1 Nov 2026. Each source supports part of the overview; availability and prices can change.
- High-Performance Compute for Agents | Beam ↗www.beam.cloud
- Pricing | Beam ↗www.beam.cloud
- Introduction - Beam ↗docs.beam.cloud
Ratings and reviews
No ratings for Beam yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- beam.cloud: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- manual_review
Last checked 1 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.