详细内容
Backed by Y Combinator
The perpetual sandbox platform
Blaxel lets you keep infinite, secure sandboxes on automatic standby while co-hosting your agents for near instant latency.
TSPY
1. Spawn sandbox
Create sandboxes & connect in 25ms.
2. Run commands
Start processes and watch events.
3. Stream logs
Stream logs live from your sandboxes.
4. Use MCP to run
Make agentic tool calls to sandboxes.
Scroll →
import { SandboxInstance } from "@blaxel/core";
const sandbox = await SandboxInstance.create()
// List files in the sandbox
const { subdir, files } = await sandbox.fs.ls("/path/to/folder");
Demo Blaxel — Tella
03:22
TRUSTED BY AI BUILDERS FROM HIGH GROWTH STARTUPS
> Truly intelligent AI agents need to run arbitrary code, but you won't get there with standard infra_
Traditional sandboxes expire in hours to days, forcing you to build complex orchestration around to maintain state, or accept a slow, cold start every session.
Plus, when agent logic runs separately from your sandbox provider, network latency gets compounded: chained executions are even slower, and teams trade performance for operational overhead.
Blaxel Sandboxes
Run AI code in secure environments that wait on standby indefinitely when not used, eliminating cold starts without complex orchestration.
Sandboxes for your AI agents
Equip your codegen agents with secure & flexible VMs to autonomously run commands and render code.
Sub-25ms latency to boot
Run AI-generated arbitrary code
Stateful even after weeks
Air-tight isolation
\* Mark 3 infrastructure only, depending on image used
8F7D6C5E
0.5 vCPU 2GB RAM. Started 27 seconds ago.
4B3A2D1C
2 vCPU 4GB RAM. Started 1 minute ago.
7A6B5C4D
8 vCPU 16GB RAM. Started 10 minutes ago.
3E2F9F8E
8 vCPU 16GB RAM. Started 10 minutes ago.
8F7D6C5E
1 vCPU 2GB RAM. Started 10 minutes ago.
Compute resources, made for both humans and agents
Run any kind of workload on agent-native VMs. Scale up and down in one command.
Ultra-low latency
Scale down to zero
Accessible either in API or MCP
Next execution in
02:34:15
Background tasks for reliable AI products
Let your agents schedule and execute thousands of background jobs at scale. From simple cron jobs to durable workflows.
Cron-based scheduling
Real-time monitoring
Automatic retries

Highly customizable gateway for your apps
Production-critical low latency and deeply programmable global gateway to federate all your agentic workloads.
Custom processor
Model routing and fallback
Private connect


Enterprise-grade security & compliance
Built with security-first architecture and certified compliance standards to meet the most stringent enterprise requirements.
SOC 2 Type II certified infrastructure
HIPAA compliant
Region support for local data residency
Explore Blaxel Sandboxes features
Blaxel Agents Cohosting
Co-locate agent APIs, MCP servers, and batch tasks directly alongside your sandboxes to eliminate network hops and ensure the lowest possible end-to-end latency.
b
Iterate. Evaluate. Deploy.
Develop and deploy agents in any framework on our serverless infrastructure.
Framework-agnostic
Python or TypeScript
Sync to GitHub to auto-deploy
No config required
Web Search MCP
Slack MCP
Private database connect MCP
MCP servers that just work
Run asynchronous tool calls on cloud-hosted MCP servers.
Pre-built servers
Custom servers
One gateway. Hundreds of AI models.
Access LLMs from all the top providers via Blaxel's AI gateway, or route to your self-hosted models.
Unified access control
Token usage rate limit
LLM routing
Model fallbacks
LLM: Parse user request
920ms
Tool: Web Search API call
150ms
LLM: Format response
380ms
Embedded LLM observability
Get a 360-view of metrics, logs, and traces for LLM applications and agent runs.
OpenTelemetry-based
Compatible with any framework
