Blaxel TypeScript SDK
Blaxel is a perpetual sandbox platform that achieves near instant latency by keeping infinite secure sandboxes on automatic standby, while co-hosting your agent logic to cut network overhead.
This repository contains Blaxel's TypeScript SDK, which lets you create and manage sandboxes and other resources on Blaxel.
Installation
# npm
npm install @blaxel/core
# yarn
yarn add @blaxel/core
# bun
bun add @blaxel/core
Authentication
The SDK authenticates with your Blaxel workspace using these sources (in priority order):
- Blaxel CLI, when logged in
- Environment variables in
.envfile (BL_WORKSPACE,BL_API_KEY) - System environment variables
- Blaxel configuration file (
~/.blaxel/config.yaml)
When developing locally, the recommended method is to just log in to your workspace with the Blaxel CLI:
bl login YOUR-WORKSPACE
This allows you to run Blaxel SDK functions that will automatically connect to your workspace without additional setup. When you deploy on Blaxel, this connection persists automatically.
When running Blaxel SDK from a remote server that is not Blaxel-hosted, we recommend using environment variables as described in the third option above.
API error types
@blaxel/core exports BlaxelErrorCode (gateway, auth and sandbox-creation
code unions), BlaxelErrorCodeValue (also unknown string codes and numeric HTTP
codes), and body types: BlaxelApiErrorBody, BlaxelActionErrorBody,
BlaxelPlatformErrorBody, BlaxelSandboxApiErrorBody.
Errors are not wrapped or changed. The SDK still throws raw control-plane
bodies, existing sandbox error classes, plain Error, or text/HTML bodies.
isBlaxelError is true only for the shapes the SDK throws for API errors
(BlaxelErrorLike): ResponseError (including SandboxGatewayError),
SandboxCreationTimeoutError, and raw control-plane error bodies that carry a
code and status. It is false for plain Errors (a network error with a code such
as ECONNRESET included), strings, raw HTML/text bodies and empty bodies. Like
is_blaxel_error in the Python SDK, it is false for plain exceptions and
strings. Unlike Python, a control-plane error with a non-JSON body (an HTML 502,
say) is thrown as a bare string here, so the guard is false for it; Python raises
UnexpectedStatus, for which is_blaxel_error is true. Use the non-throwing
accessors instead of assuming common properties:
import {
SandboxInstance, isBlaxelError, getBlaxelErrorCode,
getBlaxelErrorStatus, getBlaxelErrorMessage, getBlaxelErrorRequestId,
} from "@blaxel/core";
try {
await SandboxInstance.create({ name: "my-sandbox", region: "us-was-1" });
} catch (err) {
if (isBlaxelError(err)) {
console.error(getBlaxelErrorCode(err), getBlaxelErrorStatus(err),
getBlaxelErrorMessage(err), getBlaxelErrorRequestId(err));
}
throw err;
}
Unknown codes need a fallback. Missing fields return undefined; in particular,
raw control-plane bodies and creation-timeout errors do not retain response
headers/request IDs, and getBlaxelErrorStatus is undefined for transport
errors such as an HTTP/2 stream reset. All accessors accept unknown safely,
including values the guard rejects. POLICY_VIOLATION is reserved, not emitted today.
Usage
Sandboxes
Sandboxes are secure, instant-launching compute environments that scale to zero after inactivity and resume in under 25ms.
Base image contents: The default
blaxel/base-image:latestis Alpine Linux with Node.js 22 and git pre-installed. It does not include Python or other language runtimes. To use Python, either specifyblaxel/py-app:latestas your image (Python 3.12) or install it in the base image withapk add --no-cache python3 py3-pip.
import { SandboxInstance } from "@blaxel/core";
// Create a new sandbox
const sandbox = await SandboxInstance.createIfNotExists({
name: "my-sandbox",
image: "blaxel/base-image:latest",
memory: 4096,
region: "us-pdx-1",
ports: [{ target: 3000, protocol: "HTTP" }],
labels: { env: "dev", project: "my-project" },
ttl: "24h"
});
// Get existing sandbox
const existing = await SandboxInstance.get("my-sandbox");
// Delete sandbox (using class)
await SandboxInstance.delete("my-sandbox");
// Delete sandbox (using instance)
await existing.delete();
Listing resources
List methods return one page at a time. The result contains data, meta, and helpers for fetching more pages when you want them.
import { SandboxInstance } from "@blaxel/core";
const page = await SandboxInstance.list({
limit: 50,
sort: "createdAt:desc",
});
for (const sandbox of page.data) {
console.log(sandbox.metadata.name);
}
if (page.hasMore) {
const nextPage = await page.nextPage();
console.log(nextPage?.data);
}
// Optional auto-pagination, inspired by Stripe's SDK.
for await (const sandbox of page) {
console.log(sandbox.metadata.name);
}
const firstThousand = await page.autoPagingToArray({ limit: 1000 });
Preview URLs
Generate public preview URLs to access services running in your sandbox:
import { SandboxInstance } from "@blaxel/core";
// Get existing sandbox
const sandbox = await SandboxInstance.get("my-sandbox");
// Start a web server in the sandbox
await sandbox.process.exec({
command: "npm run dev -- --port 3000",
workingDir: "/app",
waitForPorts: [3000]
});
// Create a public preview URL
const preview = await sandbox.previews.createIfNotExists({
metadata: { name: "app-preview" },
spec: {
port: 3000,
public: true
}
});
console.log(preview.spec?.url); // https://xyz.preview.bl.run
Previews can also be private, with or without a custom prefix. When you create a private preview URL, a token is required to access the URL, passed as a request parameter or request header.
// ...
// Shorthand: preview-3000, private on creation, no prefix required.
const privatePreview = await sandbox.previews.createIfNotExists({ port: 3000 });
const token = await privatePreview.tokens.createIfExpired();
const url = new URL(privatePreview.url); // base URL, without a token
url.searchParams.set("bl_preview_token", token.value); // credential: do not log
// When no longer needed: await privatePreview.tokens.delete(token.name);
// Create a public preview URL with a custom prefix
const customPreview = await sandbox.previews.createIfNotExists({
metadata: { name: "custom-app-preview" },
spec: {
port: 3000,
prefixUrl: "my-app",
public: true
}
});
The shorthand name is constant per port within each sandbox. You can supply name
and public; custom prefixes and other advanced settings still use the full model
above. Existing previews are returned as-is, even if their port or public
setting differs: the private default only applies on creation.
createIfExpired(expiresAt?, minValidity?) defaults to a new 24-hour expiry and
one hour of minimum remaining validity (milliseconds). It reuses the latest valid,
nonempty token no later than the requested expiry, so an explicitly shorter expiry
never reuses a longer-lived credential. Token names are generated and exposed for
deletion. Concurrent callers can mint separate tokens; old tokens are not deleted.
The helpers simplify setup, not its network request count or atomicity.
Process execution
Execute and manage processes in your sandbox:
import { SandboxInstance } from "@blaxel/core";
// Get existing sandbox
const sandbox = await SandboxInstance.get("my-sandbox");
// Execute a command
const process = await sandbox.process.exec({
name: "build-process",
command: "npm run build",
workingDir: "/app",
waitForCompletion: true,
timeout: 60 // 60 seconds
});
// Kill a running process
await sandbox.process.kill("build-process");
wait() returns only a terminal API state. Temporary connection failures are retried
until maxWait; use maxWait: -1 to wait without a deadline, still cancellable with
an AbortSignal. Other errors are propagated. A timeout or cancellation stops waiting,
not the command. streamLogs().wait() also rejects when the stream fails.
Give the command a unique name before starting it so you can reconnect after a lost response without starting it twice:
const name = crypto.randomUUID();
await sandbox.process.exec({ name, command: "npm run build" });
const result = await sandbox.process.wait(name, { maxWait: 30_000 });
// After a connection error, use get(name), wait(name), or logs(name) to reconnect.
// To request termination explicitly:
await sandbox.process.kill(name); // or stop(name) for SIGTERM
await sandbox.process.wait(name);
// If the stop/kill response is lost, inspect get(name) or wait(name) before retrying.
Terminal states reflect the sandbox API's report; older servers can report stopped before the OS process exits. The SDK never automatically resends process creation.
Restart a process if it fails, up to a maximum number of restart attempts:
// ...
// Run with auto-restart on failure
process = await sandbox.process.exec({
name: "web-server",
command: "npm run dev -- --host 0.0.0.0 --port 3000",
restartOnFailure: true,
maxRestarts: 5
});
Filesystem operations
Manage files and directories within your sandbox:
import { SandboxInstance } from "@blaxel/core";
import * as fs from "fs";
// Get existing sandbox
const sandbox = await SandboxInstance.get("my-sandbox");
// Write and read text files
await sandbox.fs.write("/app/config.json", '{"key": "value"}');
const content = await sandbox.fs.read("/app/config.json");
// Write and read binary files
const binaryData = fs.readFileSync("./image.png");
await sandbox.fs.writeBinary("/app/image.png", binaryData);
const blob = await sandbox.fs.readBinary("/app/image.png");
// Create directories
await sandbox.fs.mkdir("/app/uploads");
// List files
const { subdirectories, files } = await sandbox.fs.ls("/app");
// Search for text within files
const matches = await sandbox.fs.grep("pattern", "/app", {
caseSensitive: true,
contextLines: 2,
maxResults: 5,
filePattern: "*.ts",
excludeDirs: ["node_modules"]
});
// Find files and directories matching specified patterns:
const results = await sandbox.fs.find("/app", {
type: "file",
patterns: ["*.md", "*.html"],
maxResults: 1000
});
// Watch for file changes
const handle = sandbox.fs.watch("/app", (event) => {
console.log(event.op, event.path);
}, {
withContent: true,
ignore: ["node_modules", ".git"]
});
// Close watcher
handle.close();
Drives
Pass mountDrives to mount drives on the sandbox. Drives are looked up or created while the sandbox is being created (up to 5 at a time), and each is mounted once both it and the sandbox are ready. Give driveName for a drive that already exists, or create for a new one (a named drive is reused if it already exists). New drives are created in the region the sandbox request sends (region, or BL_REGION); without one, they are created once the sandbox exists, in its region. An existing drive must be in the sandbox's region too.
import { SandboxInstance } from "@blaxel/core";
const config = { image: "blaxel/base-image:latest", region: "us-was-1" };
const sandbox = await SandboxInstance.create(config, {
mountDrives: [{ create: { name: "app-data" }, mountPath: "/mnt/data" }],
});
await sandbox.fs.write("/mnt/data/hello.txt", "hello");
// A second sandbox can mount the same drive by name.
const reader = await SandboxInstance.create(config, {
mountDrives: [{ driveName: "app-data", mountPath: "/mnt/data", readOnly: true }],
});
console.log(await reader.fs.read("/mnt/data/hello.txt")); // hello
Each entry also takes drivePath (a sub-folder of the drive). With createIfNotExists, a new drive needs a name. If anything fails, the drives this call created under a generated name (no name given) and did not mount are deleted; a drive created under a name you chose is never deleted, since a concurrent call may be using it, and is listed in createdDrives instead. If a drive can't be set up or mounted, create throws SandboxDriveSetupError with the sandbox, the driveNames left in place, the createdDrives among them that this call created (or may have created, if a response was lost) and the original cause; the sandbox and the mounts made so far are kept. If the sandbox itself can't be created, its error is thrown, wrapped in a SandboxDriveSetupError without a sandbox only if a drive created for it is left in place.
Volumes
Persist data by attaching and using volumes:
import { VolumeInstance, SandboxInstance } from "@blaxel/core";
// Create a volume
const volume = await VolumeInstance.createIfNotExists({
name: "my-volume",
size: 1024, // MB
region: "us-pdx-1",
labels: { env: "test", project: "12345" }
});
// Attach volume to sandbox
const sandbox = await SandboxInstance.createIfNotExists({
name: "my-sandbox",
image: "blaxel/base-image:latest",
volumes: [
{ name: "my-volume", mountPath: "/data", readOnly: false }
]
});
// List first page of volumes
const volumes = await VolumeInstance.list();
for (const volume of volumes.data) {
console.log(volume.name);
}
// Delete volume (using class)
await VolumeInstance.delete("my-volume");
// Delete volume (using instance)
await volume.delete();
Batch jobs
Blaxel lets you support agentic workflows by offloading asynchronous batch processing tasks to its scalable infrastructure, where they can run in parallel. Jobs can run multiple times within a single execution and accept optional input parameters.
import { blJob } from "@blaxel/core";
// Create and run a job execution
const job = blJob("job-name");
const executionId = await job.createExecution({
tasks: [
{ name: "John" },
{ name: "Jane" },
{ name: "Bob" }
]
});
// Get execution status
// Returns: "pending" | "running" | "completed" | "failed"
const status = await job.getExecutionStatus(executionId);
// Get execution details
const execution = await job.getExecution(executionId);
console.log(execution.status, execution.metadata);
// Wait for completion
try {
const result = await job.waitForExecution(executionId, {
maxWait: 300000, // 5 minutes (milliseconds)
interval: 2000, // Poll every 2 seconds (milliseconds)
});
console.log(`Completed: ${result.status}`);
} catch (error) {
console.log(`Timeout: ${error.message}`);
}
// List first page of executions
const executions = await job.listExecutions();
for (const execution of executions.data) {
console.log(execution.status);
}
// Delete an execution
await job.deleteExecution(executionId);
Framework integrations
Blaxel provides additional packages for framework-specific integrations and telemetry:
@blaxel/telemetry- OpenTelemetry instrumentation@blaxel/eve-sandbox- Durable Blaxel sandbox backend for eve@blaxel/vercel- Vercel AI SDK integration@blaxel/llamaindex- LlamaIndex integration@blaxel/langgraph- LangGraph integration@blaxel/mastra- Mastra integration
Model use
Blaxel acts as a unified gateway for model APIs, centralizing access credentials, tracing and telemetry. You can integrate with any model API provider, or deploy your own custom model. When a model is deployed on Blaxel, a global API endpoint is also created to call it.
The SDK includes a helper function that creates a reference to a model deployed on Blaxel and returns a framework-specific model client that routes API calls through Blaxel's unified gateway.
import { blModel } from "@blaxel/core";
// With Vercel AI SDK
import { blModel } from "@blaxel/vercel";
const model = await blModel("gpt-5-mini");
// With LangChain
import { blModel } from "@blaxel/langgraph";
const model = await blModel("gpt-5-mini");
// With LlamaIndex
import { blModel } from "@blaxel/llamaindex";
const model = await blModel("gpt-5-mini");
// With Mastra
import { blModel } from "@blaxel/mastra";
const model = await blModel("gpt-5-mini");
MCP tool use
Blaxel lets you deploy and host Model Context Protocol (MCP) servers, accessible at a global endpoint over streamable HTTP.
The SDK includes a helper function that retrieves and returns tool definitions from a Blaxel-hosted MCP server in the format required by specific frameworks.
// With Vercel AI SDK
import { blTools } from "@blaxel/vercel";
const tools = await blTools(['sandbox/my-sandbox'])
// With Mastra
import { blTools } from "@blaxel/mastra";
const tools = await blTools(['sandbox/my-sandbox'])
// With LlamaIndex
import { blTools } from "@blaxel/llamaindex";
const tools = await blTools(['sandbox/my-sandbox'])
// With LangChain
import { blTools } from "@blaxel/langgraph";
const tools = await blTools(['sandbox/my-sandbox'])
Here is an example of retrieving tool definitions from a Blaxel sandbox's MCP server in the format required by Vercel's AI SDK:
import { SandboxInstance } from "@blaxel/core";
import { blTools } from "@blaxel/vercel";
// Create a new sandbox
const sandbox = await SandboxInstance.createIfNotExists({
name: "my-sandbox",
image: "blaxel/base-image:latest",
memory: 4096,
region: "us-pdx-1",
ttl: "24h"
});
// Get sandbox tools
const tools = await blTools(['sandbox/my-sandbox'])
Telemetry
Instrumentation happens automatically when workloads run on Blaxel.
Enable automatic telemetry by importing the @blaxel/telemetry package:
import "@blaxel/telemetry";
Data collection
Error tracking
The SDK includes opt-in error tracking for unhandled exceptions whose first non-runtime frame belongs to the installed @blaxel/core package. Events contain an allowlisted error type, a fixed message, SDK-owned relative stack frames, the SDK version and commit, and the dev or prod environment. Raw error messages, response bodies, workspace and resource identifiers, absolute paths, and application stack frames are not collected. Caught errors, including errors logged with console.error, are not reported. If a browser bundler removes the package boundary from stack frames, error tracking stays disabled rather than attributing application errors to the SDK.
Error tracking is off by default since v0.2.76. To explicitly disable it in older versions:
export DO_NOT_TRACK=1
Or add to ~/.blaxel/config.yaml:
tracking: false
Where both settings exist, the DO_NOT_TRACK variable takes precedence.
Telemetry
Telemetry, delivered via OpenTelemetry, is controlled by the BL_ENABLE_OPENTELEMETRY environment variable.
When you deploy an agent to Blaxel, the platform automatically injects BL_ENABLE_OPENTELEMETRY=true into the environment.
When developing locally, this environment variable is not set and therefore defaults to false.
To explicitly disable telemetry, override the variable in your Blaxel deployment:
export BL_ENABLE_OPENTELEMETRY=false
For more information, refer to our documentation.
Requirements
- Node.js v18 or later
Contributing
Contributions are welcome! Please feel free to submit a pull request.
License
This project is licensed under the MIT License. See the LICENSE file for details.
Image repository and tag pagination
The SDK defaults to API version 2026-09-22. listImages() now returns
{ data, meta } with repository summaries (metadata and spec.size /
spec.tagCount), rather than an array containing every tag. Fetch tags separately
with listImageTags() or use the lazy helpers:
import { ImageInstance } from "@blaxel/core";
const images = await ImageInstance.list({ limit: 20, sort: "name:asc", q: "my-" });
for await (const image of images) {
console.log(image.metadata.name, image.spec.tagCount);
}
const tags = await ImageInstance.listTags("sandbox", "my-image", {
limit: 20,
sort: "name:desc",
// sourceWorkspace: "owner-workspace", // For an account-shared image
});
const first100Tags = await tags.autoPagingToArray({ limit: 100 });
Each helper fetches one page initially. nextPage(), async iteration and
autoPagingToArray() fetch subsequent pages on demand. Tag search supports q
(prefix) or name (exact), never both; tag sorting is by name only.
BL_API_VERSION remains available for older contracts through low-level API calls;
image pagination and the typed summaries described above require 2026-09-22
or later.