Claude Code · Blaxel Sandbox

Claude works, waits, then works again.

Its infrastructure should too. Blaxel keeps the full working state ready through standby, resumes it in 25ms, and drops active-compute cost to $0 while Claude waits.

Up to $200 in credits · No credit card required

Teams building with Blaxel

Webflow logoPloy logoStrapi logoSapiom logo
SOC 2 CertifiedHIPAA CompliantISO 27001 Certified

Claude + Blaxel

Claude does the work. Blaxel keeps its workspace ready.

Run Claude Code or Claude Agent SDK on Blaxel, or connect Claude to a separate application sandbox over MCP. The work stays in a persistent execution environment between turns and reviews.

ClaudeAgent layer

Claude

Use Claude Code for interactive development or Claude Agent SDK to build agents with Claude's models, tools, permissions, and MCP connections.

  • Claude Code and Agent SDK
  • Models, tools, and permissions
  • Agent behavior and MCP use

Claude execution paths

hosting · MCP · managed agents

BlaxelExecution layer

Blaxel Sandbox

Give Claude an isolated computer where it can operate a real codebase, keep useful state, run services, and expose the result for review.

  • Commands, files, and packages
  • $0 active compute on standby
  • 25ms resume with state intact
  • Network and secret controls in preview

Together, Claude can

Execute

Run shell commands, tests, builds, and generated code inside an isolated microVM.

Work

Read and edit a real codebase, install packages, and move artifacts in and out.

Preview

Start an application and open a live URL so people can review what Claude built.

Continue

Keep its filesystem, memory, and processes intact on standby, then resume the same work in 25ms.

See Claude return to the same workspace.

Up to $200 in credits · No credit card required

Run Claude in a sandbox

Claude Code over sandbox MCP

Connect Claude to a live workspace.

Every Blaxel Sandbox exposes a built-in MCP server. Add that endpoint to Claude Code, then let Claude operate the sandbox through tools for files, processes, code search, and previews.

  1. 01

    Start a sandbox for the work

    Choose an image for the application, repository, dependencies, and services Claude will operate.

  2. 02

    Add the sandbox MCP endpoint

    Connect Claude Code to the sandbox's built-in MCP server with your Blaxel API key.

  3. 03

    Send Claude a real task

    Let Claude inspect files, run commands, make changes, and show the result through a preview URL.

terminal
Shell
$bl create sandbox app-runtime --image blaxel/nextjs:latest
1export BLAXEL_API_KEY="..."2export SANDBOX_MCP_URL="https://<sandbox-url>/mcp"3 4claude mcp add --transport http sandbox   "$SANDBOX_MCP_URL"   --header "Authorization: Bearer $BLAXEL_API_KEY"5 6claude mcp list7claude

Claude Code · Blaxel sandbox MCP

Open the full Claude Code guide

Persistence across Claude workflows

Claude should not rebuild its workspace between turns.

Blaxel keeps the full sandbox state on standby between actions. Active compute costs $0 while it waits, and the same files, memory, and processes resume in 25ms.

$0 active compute on standby

Keep the work without keeping compute active.

Move the sandbox to standby between Claude turns or human reviews. The working state stays ready while active-compute billing stops.

  • $0 active compute on standby
  • Filesystem, memory, and processes retained
  • 25ms resume

Claude Code

Start Claude inside a ready-to-work sandbox.

Use Blaxel's Claude Code image, connect it to an application sandbox over MCP, and give it a live codebase instead of a disposable shell.

  • Claude Code available on PATH
  • Sandbox MCP over streamable HTTP
  • Private application previews

Claude Agent SDK

Deploy the agent or connect it to sandbox tools.

Deploy Claude Agent SDK as an API on Blaxel Agents Hosting. When it needs a separate execution environment, connect it to the sandbox's built-in MCP endpoint.

  • Blaxel Agents Hosting
  • Standard MCP connection
  • Process, filesystem, and code tools

Managed Agents beta path

Keep the responsibility boundary clear.

With self-hosted Claude Managed Agents, Anthropic runs the agent loop and work queue. The Blaxel reference deployment creates worker sandboxes for tool execution.

  • Anthropic-managed agent loop
  • Blaxel worker sandboxes
  • Webhook-driven orchestration

Build with Claude

Choose the Claude workflow that fits the job.

Use the same sandbox platform for interactive coding, SDK-built agents, and the self-hosted execution contract in Claude Managed Agents.

Claude

Hosted and MCP execution

Claude Code + Agent SDK

Run Claude Code in a sandbox, deploy Claude Agent SDK on Agents Hosting, or connect either workflow to a separate sandbox through MCP.

Claude

Self-hosted beta path

Claude Managed Agents

Keep the agent loop and session orchestration on Anthropic while the Blaxel reference implementation runs sandbox tool execution.

Start with Claude Code or Claude Agent SDK.

Up to $200 in credits · No credit card required

Run Claude in a sandbox

Claude sandbox

Questions before you run Claude in a sandbox.

What can Claude Code do in a Blaxel Sandbox?

Claude Code can use the sandbox's MCP tools to inspect and edit files, search a codebase, execute commands, manage processes, and open application previews. The work happens inside the Blaxel execution environment you connect.

Does Claude Agent SDK require MCP on Blaxel?

No. You can deploy a Claude Agent SDK agent directly on Blaxel Agents Hosting. Use MCP when the agent needs to operate a separate sandbox through process, filesystem, code, port, and preview tools.

Does the Claude workspace persist between interactions?

Yes. A Blaxel application sandbox can move to standby while retaining its filesystem, memory, and running processes, then resume the same working state in 25ms when Claude or a person returns. This lifecycle is separate from the short-lived worker pattern in the Managed Agents reference deployment.

What does $0 active compute on standby mean?

When the sandbox moves to standby, active-compute billing stops. Snapshot storage and any attached volume storage continue to apply while Blaxel preserves the working state for its next 25ms resume.

How does Blaxel work with Claude Managed Agents?

Claude Managed Agents keeps the agent loop, session orchestration, and work queue on Anthropic. In the self-hosted path, the Blaxel reference implementation receives session events and creates worker sandboxes that run tool calls, file operations, and code. Managed Agents is currently a beta Anthropic product.

How do I control network access and secrets?

Domain filtering and proxy secret injection are available in public preview. Dedicated egress IPs are available in private preview when an external service requires a stable allowlisted identity.

Claude + Blaxel

Give Claude a workspace it can return to.

Keep files, memory, and processes ready between turns. Active compute costs $0 on standby, and the same sandbox resumes in 25ms.

Run Claude in a sandbox