@blinkdotnew/cli

The Blink platform CLI. Deploy apps, query databases, generate AI content, manage storage, and control Blink Claw agents — all from your terminal or agent scripts.

npm install -g @blinkdotnew/cli
blink --help

Quick Start

# Install
npm install -g @blinkdotnew/cli

# Authenticate (enter your blnk_ak_... API key from blink.new → Settings → API Keys)
blink login --interactive

# See full context: agent + project + auth
blink status

# Manage Claw agents
blink agent list
eval $(blink agent use clw_xxx --export)
blink secrets set GITHUB_TOKEN ghp_xxx
blink secrets list

# Deploy an app
npm run build && blink deploy ./dist --prod

# Query your database
blink db query "SELECT count(*) FROM users"

# Generate AI content
blink ai image "a glowing blink logo on dark background"
blink ai text "Summarize this article: ..."

# Scrape websites
blink scrape https://lovable.dev --extract "pricing tiers and costs"
blink scrape https://example.com --text

Authentication

The CLI resolves credentials in this order (highest priority first):

| Source | How | |---|---| | --token <key> flag | One-off override per command | | BLINK_API_KEY env var | CI/CD and Claw agents (set automatically) | | ~/.config/blink/config.toml | Stored after blink login |

# Interactive login (saves to ~/.config/blink/config.toml)
blink login --interactive

# One-off with token flag
blink deploy --token blnk_ak_xxx ./dist

# CI / GitHub Actions
BLINK_API_KEY=blnk_ak_xxx blink deploy ./dist --prod

Your API key lives at blink.new → Settings → API Keys (blnk_ak_...).

Multiple accounts

# ~/.config/blink/config.toml
[default]
api_key = "blnk_ak_personal"

[work]
api_key = "blnk_ak_work"

blink --profile work deploy ./dist

Commands

blink deploy — Deploy to Blink hosting

blink deploy                          # Deploy CWD, use linked project
blink deploy ./dist                   # Deploy specific build dir
blink deploy proj_xxx ./dist          # Explicit project + dir
blink deploy ./dist --prod            # Production (live domain)
blink deploy ./dist --preview         # Preview URL (default)
blink deploy ./dist --name "v2 beta"  # Label this deployment

blink deployments                     # List deployments
blink rollback                        # Rollback production to previous deploy

The CLI packages your pre-built output directory and uploads it. Run your build first:

npm run build && blink deploy ./dist --prod

Preview deploys get a URL like preview-a1b2c3d4.sites.blink.new.
Production deploys go to your custom domain or {id}.sites.blink.new.

blink project — Project management

blink project list                    # List all workspace projects
blink project create "My App"         # Create a new project
blink project delete proj_xxx         # Delete a project (confirms first)

blink link                            # Link current dir to a project (interactive picker)
blink link proj_xxx                   # Link to specific project
blink unlink                          # Remove .blink/project.json
blink status                          # Show linked project + auth source

Once linked, commands that need a project ID use .blink/project.json automatically.

blink use — Set active project for a shell session

blink use proj_xxx                           # Prints the export command
eval $(blink use proj_xxx --export)          # Actually sets it in your shell
export BLINK_ACTIVE_PROJECT=proj_xxx         # Or set manually

After this, omit proj_xxx from any command — it's inferred from the env var.

blink db — Database (Turso/SQLite per project)

# Query (two forms)
blink db query "SELECT * FROM users LIMIT 10"
blink db query proj_xxx "SELECT * FROM users LIMIT 10"

# Execute a SQL file
blink db exec schema.sql
blink db exec proj_xxx migrations/001.sql

# List tables
blink db list

# List rows in a table
blink db list users
blink db list proj_xxx users --limit 50

# Machine-readable output
blink db query "SELECT id, email FROM users" --json

blink storage — File storage

# Upload
blink storage upload ./logo.png
blink storage upload proj_xxx ./logo.png --path images/logo.png

# List
blink storage list
blink storage list images/

# Download
blink storage download images/logo.png ./local-logo.png

# Get public URL
blink storage url images/logo.png

# Delete
blink storage delete images/old-logo.png

blink ai — AI generation

# Text
blink ai text "Summarize this in 3 bullets: ..."
blink ai text "Write a product description" --model anthropic/claude-sonnet-4.5

# Image
blink ai image "a futuristic city at sunset"
blink ai image "a futuristic city" --model fal-ai/nano-banana-pro --n 4
blink ai image "a city at sunset" --output ./city.jpg

# Image editing
blink ai image-edit "make it night time" https://example.com/city.jpg

# Video (text-to-video)
blink ai video "a drone flyover of a neon city"
blink ai video "timelapse of clouds" --model fal-ai/veo3.1 --duration 10s --aspect 9:16

# Video (image-to-video) — local file or URL
blink ai animate "make it come alive" ./photo.jpg
blink ai animate "pan slowly to the right" https://example.com/photo.jpg

# Speech (text-to-speech)
blink ai speech "Hello, welcome to Blink."
blink ai speech "Hello world" --voice nova --output ./greeting.mp3
# Voices: alloy, echo, fable, onyx, nova, shimmer

# Transcription
blink ai transcribe ./meeting.mp3
blink ai transcribe https://example.com/audio.mp3 --language en

blink fetch, blink search, blink scrape — Web & data

# Fetch any URL via Blink proxy (handles CORS, auth headers)
blink fetch https://api.github.com/users/octocat
blink fetch https://api.example.com/data --method POST --body '{"key":"val"}'
blink fetch https://api.example.com --header "X-API-Key: secret"

# Web search
blink search "latest AI news"
blink search "React Server Components" --count 10 --json

# Scrape web pages
blink scrape https://example.com                              # Raw response
blink scrape https://example.com --text                      # Clean text (strips HTML)
blink scrape https://example.com --extract "all prices"      # AI-extract specific data
blink scrape https://news.ycombinator.com --extract "top 10 story titles and URLs"
blink scrape https://example.com --extract "contact email" --json

No project key needed — blink scrape uses your workspace key only.

• --text — strips all HTML tags, returns readable text
• --extract <instructions> — uses AI (Gemini Flash) to extract exactly what you ask for
• Combine with --json for { url, content } or { url, extracted } output

blink realtime — Pub/sub

# Publish an event to a channel
blink realtime publish updates '{"type":"refresh"}'
blink realtime publish proj_xxx updates '{"type":"refresh","data":{"count":42}}'

blink rag — Knowledge base / RAG

# Search
blink rag search "how does billing work"
blink rag search "billing" --ai            # AI-enhanced search
blink rag search proj_xxx "billing" --limit 10

# Upload a document
blink rag upload ./docs/faq.md
blink rag upload proj_xxx ./docs/faq.md --collection coll_xxx

# List collections
blink rag collections

blink notify — Email

blink notify email [email protected] "Welcome!" "Thanks for signing up."
blink notify email [email protected] "Newsletter" --file ./email.html
blink notify email proj_xxx [email protected] "Subject" "Body"

blink connector — OAuth connectors

# Execute a connector action (Notion, Google, Slack, HubSpot, etc.)
blink connector exec notion search_pages '{"query":"meeting notes"}'
blink connector exec google_calendar list_events '{}'
blink connector exec slack post_message '{"channel":"C123","text":"Hello"}' --method POST
blink connector exec notion create_page '{"title":"New Page"}' --account acct_xxx

Global Flags

Available on every command:

| Flag | Description | |---|---| | --token <key> | Override API key for this command only | | --json | Output raw JSON (no colors, no spinners — great for scripting) | | --yes | Skip confirmation prompts | | --profile <name> | Use a named profile from config.toml | | --debug | Print full request/response details | | --version | Print CLI version | | --help | Show help for any command |

Scripting & CI/CD

The --json flag makes every command machine-readable:

# Get deployment URL from a deploy
URL=$(blink deploy ./dist --prod --json | jq -r '.url')
echo "Deployed to: $URL"

# Query database and pipe to jq
blink db query "SELECT id, email FROM users" --json | jq '.[].email'

# Generate image and get URL
IMG=$(blink ai image "a logo" --json | jq -r '.url')
blink storage upload ./banner.png --json | jq -r '.url'

GitHub Actions example

- name: Deploy to Blink
  env:
    BLINK_API_KEY: ${{ secrets.BLINK_API_KEY }}
  run: |
    npm install -g @blinkdotnew/cli
    blink deploy ./dist --prod --json

Config Files

| File | Purpose | |---|---| | ~/.config/blink/config.toml | Auth tokens, workspace ID, named profiles | | .blink/project.json | Per-directory project link (like .vercel/project.json) |

.blink/project.json is created by blink link — add it to .gitignore.

Blink Claw Agents

The CLI is pre-installed in every Blink Claw agent (Fly.io Firecracker VM). These env vars are already injected — no login needed:

BLINK_API_KEY    auth token (workspace-scoped)
BLINK_AGENT_ID   this agent's ID — auto-used by blink agent/secrets commands
BLINK_APIS_URL   https://core.blink.new
BLINK_APP_URL    https://blink.new

Agents use the CLI directly in skill scripts:

# In openclaw/skills/blink-image/scripts/generate.sh
blink ai image "$PROMPT" --model "$MODEL" --json

blink agent — Manage agents

blink agent list                           # List all agents in workspace
blink agent use clw_xxx                    # Show how to set active agent
eval $(blink agent use clw_xxx --export)   # Set active agent for this session
blink agent status                         # Show current agent details
blink agent status clw_xxx                 # Show specific agent details

On Claw machines, BLINK_AGENT_ID is already set — blink agent status works with zero config.

blink secrets — Agent encrypted vault

# All commands auto-use BLINK_AGENT_ID on Claw machines
blink secrets list                         # List secret key names (values never shown)
blink secrets set GITHUB_TOKEN ghp_xxx     # Add or update a secret
blink secrets delete OLD_KEY               # Remove a secret (--yes to skip prompt)

# Cross-agent management (agent manager pattern)
blink secrets list --agent clw_other       # Another agent's keys
blink secrets set --agent clw_other OPENAI_KEY sk-xxx   # Set on another agent

Secrets are stored encrypted and available as $KEY_NAME in all agent shell commands.

Agent ID resolution for blink agent and blink secrets:

1. --agent <id> flag
2. BLINK_AGENT_ID env var ← always set on Claw Fly machines
3. BLINK_ACTIVE_AGENT env var ← from eval $(blink agent use clw_xxx --export)

blink status — Full context at a glance

$ BLINK_AGENT_ID=clw_xxx blink status

Agent    clw_xxx           (BLINK_AGENT_ID env)
Project  proj_yyy          (.blink/project.json)
Auth     ~/.config/blink/config.toml

Links

• npm: https://www.npmjs.com/package/@blinkdotnew/cli
• Blink platform: https://blink.new
• Docs: https://blink.new/docs
• Source: blink-sdk/packages/cli/ in the blink-new/blink-sdk repo