DeployBeacon AI · Invitation-only private beta

Use the same publishing workflow through REST, CLI or MCP.

DeployBeacon's dashboard, API, downloadable CLI and compatible MCP clients share projects, inspections and release history. Use the route that fits your tools while keeping credentials private and publication explicit.

REST API

The live OpenAPI contract lists implemented endpoints and authentication requirements. The production API base is https://deploybeacon.ai/api/v1. Writes use a bearer token with appropriate scopes. Publishing requires an Idempotency-Key header and explicit confirmation; keep a key for a single intended operation when recovering from an uncertain response.

Read the returned state and evidence before proceeding. An inspection identifier is not proof that a release is live. The result of publication includes the project and release outcome.

Read recorded domain status through MCP

The domain_status tool accepts a project identifier and optional pagination cursor and limit. It requires both projects:read and domains:read, and returns stored connection states, check timestamps and manual DNS instructions. Its check mode is recorded; it does not run a fresh probe, mutate the connection or act at a DNS provider. OAuth consent must explicitly request domain-read access. Existing grants keep their original scopes.

Downloadable CLI

Download the CLI from Integrations in your invited workspace and use Node.js 22 or later. Create a scoped token in Access & security and set it privately in your terminal environment. The CLI is downloaded from the product; this guide does not instruct you to install an unclaimed npm package.

# Set these privately in your terminal
DEPLOYBEACON_URL=https://deploybeacon.ai
DEPLOYBEACON_TOKEN=<your-scoped-token>

node deploybeacon.mjs projects
node deploybeacon.mjs inspect PROJECT_ID built-site.zip
node deploybeacon.mjs deploy INSPECTION_ID --confirm
node deploybeacon.mjs release RELEASE_ID
node deploybeacon.mjs rollback RELEASE_ID --confirm

Syntax for setting an environment variable differs between shells. Use your terminal's private environment mechanism and avoid recording tokens in command transcripts.

MCP connection

Compatible clients connect to https://deploybeacon.ai/mcp using Streamable HTTP. OAuth public clients use PKCE S256 and send the MCP endpoint as the resource during sign-in and token exchange. Review the workspace and permissions on the consent screen. Token-based clients send Authorization: Bearer with a scoped token.

Current tools can list projects, submit small built websites, read release evidence, propose publication or restore and execute approved proposals. Client-specific compatibility still needs validation; endpoint support is not a certification that every AI client works.

Direct small-file inspection

The files_inspect MCP tool accepts a project identifier, confirm: true and a files array. Each file has a relative path, content and optional utf8 or base64 encoding. The limits are 50 files, 512 KiB of decoded content and a 1 MiB request. The REST equivalent is POST /api/v1/projects/PROJECT_ID/file-imports.

This prepares a private inspection. Read its findings and create a preview or publication proposal afterward. Larger built websites use ZIP upload.

Connect without account access

A coding assistant can simply produce a ZIP for you to inspect in the dashboard. The current public handoff instructions and versioned handoff instructions describe this route. Reading documentation does not authorize an assistant to publish or access an account.

Explore the product and guides

Access is invitation-only. No payments are collected during the private beta. Sign in to your invited workspace.