@headlessly/cli

Agent-first CLI for headless.ly. Auto-provisions on first use. Structured JSON output. Zero config.

npx @headlessly/cli search Contact --json

No signup. No API key. The CLI auto-provisions an anonymous tenant on first run, saves credentials, and you're operating in seconds.

Install

npm install -g @headlessly/cli

Or run directly:

npx @headlessly/cli search Contact --json

Quick Start

Just run a command. The CLI auto-provisions a tenant on first use:

# First run — auto-provisions and saves credentials
headlessly search Contact --json
# => Auto-provisioned tenant: anon_abc123
# => []

# Create data
headlessly do create Contact --data '{"name":"Alice","email":"alice@example.com","stage":"Lead"}' --json

# Search it back
headlessly search Contact --json --fields name,stage

# See the CDC events
headlessly fetch events --json --limit 5

# Claim the tenant permanently via GitHub commit
# https://id.org.ai/claim/clm_...

search

Find entities with filters, sorting, and field masking.

headlessly search Contact --json
headlessly search Contact --filter stage=Lead --limit 10 --json
headlessly search Deal --filter "value>10000" --sort value:desc --json
headlessly search Contact --fields name,email,stage --json
headlessly search --query "alice" --json
headlessly search Contact --count
headlessly search Contact --output csv

fetch

Get a specific entity, schema, or events.

# Entity by ID
headlessly fetch Contact contact_fX9bL5nRd --json
headlessly fetch Contact contact_fX9bL5nRd --fields name,email,stage --json
headlessly fetch Contact contact_fX9bL5nRd --include deals --json

# Schema
headlessly fetch schema
headlessly fetch schema Contact

# Events (ClickHouse CDC)
headlessly fetch events --json --limit 10
headlessly fetch events --type Contact --since 2026-01-01 --json
headlessly fetch events --fields id,type,event,ts --json

# Entity history
headlessly fetch Contact contact_fX9bL5nRd --history --json

do

Create, update, delete entities, or execute custom verbs.

# Create
headlessly do create Contact --data '{"name":"Alice","stage":"Lead"}' --json

# Update
headlessly do update Contact contact_fX9bL5nRd --data '{"stage":"Qualified"}' --json

# Delete
headlessly do delete Contact contact_fX9bL5nRd --json

# Custom verbs
headlessly do qualify Contact contact_fX9bL5nRd --json
headlessly do close Deal deal_k7TmPvQx --json

# Dry run — preview without executing
headlessly do create Contact --data '{"name":"Test"}' --dry-run --json

Individual fields can also be passed as flags: --name Alice --stage Lead.

Agent-First Design

Structured output: --json on any command. JSON on stdout, status messages on stderr.

Field masking: --fields name,stage keeps context windows small.

Dry run: --dry-run previews mutations without executing.

Exit codes: 0 (success), 1 (usage), 2 (auth), 3 (not found), 4 (server).

Structured errors:

{"error":{"code":"VALIDATION","message":"Field 'stage' must be one of: Lead, Qualified, Customer"}}

Commands

Command	Description
`search [type]`	Search entities with filters, sorting, field masking
`fetch <type> <id>`	Fetch entity, schema, or events
`do <action> [type] [id]`	Create, update, delete, or execute custom verbs
`login`	Configure remote endpoint and credentials
`status`	Show connection status and registered entities
`mcp`	Start MCP server for AI agent connections
`help`	Show usage information
`--version`	Show version

Configuration

The CLI reads configuration from three sources (in order of precedence):

Environment variables: HEADLESSLY_ENDPOINT, HEADLESSLY_API_KEY or HEADLESSLY_TOKEN, HEADLESSLY_TENANT
Config file: ~/.headlessly/config.json (auto-created on first use, or by headlessly login)
Auto-provision: If no config exists, provisions an anonymous tenant automatically

# Explicit login
headlessly login --tenant my-startup --api-key hly_abc123

# Or via environment
export HEADLESSLY_API_KEY=hly_abc123
export HEADLESSLY_ENDPOINT=https://api.headless.ly/~my-startup
headlessly search Contact --json

# Or just run it — auto-provisions on first use
headlessly search Contact --json

MCP Server

Start a local MCP server for AI agent connections:

headlessly mcp

Exposes three tools: search, fetch, do.

License

MIT