anysite-cli 0.3.3

CLI for Anysite API - web data extraction for humans and AI agents

Anysite CLI

A command-line tool designed for AI agents to collect, analyze, and store data from the web and your databases — with full support for humans too.

Agent-native protocol. Auto-detects pipes and subprocesses, switches to structured JSON output. Discovery payload on first run, machine-readable exit codes, error codes with retryable flag and suggestions, next-step hints on every command. Zero configuration for agents — just call the binary.

Self-describing API. 118+ endpoints with anysite describe: input parameters, output fields with nested object/array expansion, dot-notation paths for dependency chains. An agent discovers the schema, plans collection, and executes — no documentation lookup needed.

Declarative data pipelines. Define multi-source collection workflows in YAML: dependency chains between sources, union merges, incremental collection that skips already-fetched data, per-source transforms and exports, automatic topological execution. One anysite dataset collect replaces hundreds of lines of scripting.

LLM analysis without burning tokens. Offload enrichment, classification, summarization, and deduplication to cheaper LLMs (OpenAI, Anthropic). Results are cached in SQLite — repeat runs cost nothing. Agents keep their context window for reasoning, not data crunching.

Database-ready output. Auto-infer schemas from JSON, create tables, and load into SQLite, PostgreSQL, or ClickHouse with a single command. Foreign keys are linked automatically via provenance tracking. Diff-based incremental sync keeps your database up to date without full reloads.

Database discovery. Connect any SQLite, PostgreSQL, or ClickHouse database and auto-discover its structure: tables, columns, types, indexes, foreign keys, row counts, and sample data. Optionally enrich with LLM-generated descriptions — table summaries, column semantics, implicit relationships. Saved catalogs give agents instant context about your data without manual documentation.

Supports LinkedIn (profiles, companies, jobs, Sales Navigator, email lookup), Instagram (profiles, posts, reels, comments), Twitter/X, Reddit, YouTube (channels, videos, subtitles), Y Combinator, SEC EDGAR, GitHub, Amazon, Google News, Trustpilot, TripAdvisor, Hacker News, web page parsing, and 60+ more sources via the Anysite API.

Installation

One-line install (macOS / Linux — no Python or Homebrew required, all features included):

curl -fsSL https://raw.githubusercontent.com/anysiteio/anysite-cli/main/install.sh | bash

Homebrew:

brew tap anysiteio/cli https://github.com/anysiteio/anysite-cli
brew install anysite

pip (if you already have Python 3.11+):

pip install anysite-cli

Optional extras:

pip install "anysite-cli[data]"        # DuckDB + PyArrow for dataset pipelines
pip install "anysite-cli[postgres]"    # PostgreSQL support
pip install "anysite-cli[clickhouse]"  # ClickHouse support
pip install "anysite-cli[all]"         # All optional dependencies

Authentication

Anysite account — required for web data collection. Sign up for a free trial at app.anysite.io.

Option 1: Log in via browser (recommended)

anysite auth login

Opens your browser for OAuth2 authentication. After approval, the CLI automatically receives and stores your access token (valid for 30 days).

Option 2: Set API key manually

anysite config set api_key YOUR_API_KEY

Or via environment variable:

export ANYSITE_API_KEY=sk-xxxxx

Manage your session:

anysite auth status    # Check current authentication
anysite auth logout    # Remove stored OAuth token

OpenAI or Anthropic API key — optional, for LLM-powered analysis (enrichment, classification, summarization, deduplication).

anysite llm setup

Database features (db, dataset) work without any API keys.

Agent Protocol

Anysite CLI is agent-first: it auto-detects when stdout is not a TTY (pipe, subprocess) and switches all output to structured JSON. No flags needed.

Discovery

Run anysite with no arguments in a pipe to get a full discovery payload:

anysite | jq '.result'

Returns:

• commands — all available commands with descriptions and subcommands
• agent_protocol — how auto-JSON, --json, --human, --non-interactive work
• output_schema — success/error envelope format
• exit_codes — machine-readable exit code meanings
• installed_extras — which optional packages are available (data, llm, postgres)

Discover API endpoints:

anysite describe                          # list all 118+ endpoints
anysite describe --search "company"       # search by keyword
anysite describe /api/linkedin/company    # input params + output fields with nested expansion

Nested fields are expanded with dot-notation:

Output fields (15):
    name                           string
    urn                            object
      .type                        string
      .value                       string
    experience                     array[object]
      .title                       string
      .company_urn                 string

Use these paths in --fields, dependency.field, and db_load.fields.

JSON Envelope

Every command in pipe mode returns a JSON envelope:

Success:

{
  "ok": true,
  "result": { ... },
  "hints": [{"action": "Next step", "command": "anysite ..."}],
  "meta": {"version": "0.2.0", "command": "anysite db add"}
}

Error:

{
  "ok": false,
  "error": {
    "code": "AUTH_FAILED",
    "message": "Authentication failed",
    "retryable": false,
    "suggestions": ["Set API key: anysite config set api_key <key>"]
  },
  "meta": {"version": "0.2.0"}
}

Check ok for success/failure. Use error.code for programmatic handling. Use error.retryable to decide whether to retry. Use error.suggestions and hints for next steps.

Exit Codes

Code	Meaning
0	Success
1	General error
2	Usage error (invalid args, missing params)
3	Authentication failed
4	Resource not found
5	Network / timeout / rate limit

Output Mode Flags

Context	Default	Override
Pipe / subprocess (no TTY)	JSON envelope	--human to force human text
Terminal (TTY)	Human-readable Rich text	--json to force JSON envelope

--non-interactive disables interactive prompts (auto-enabled when stdin is not a TTY).

Hints

Every command returns next-step hints — in JSON (hints array) and human mode (dim text on stderr). Agents discover follow-up commands without consulting documentation.

Built-in Guide

anysite dataset guide                        # full YAML config reference
anysite dataset guide --section sources      # specific section
anysite dataset guide --example advanced     # complete example config
anysite dataset guide --json                 # structured JSON for agents

Quick Start

1. Configure your API key

anysite config set api_key sk-xxxxx

Or set environment variable:

export ANYSITE_API_KEY=sk-xxxxx

2. Update the schema cache

anysite schema update

3. Make your first request

anysite api /api/linkedin/user user=satyanadella

The api Command

A single universal command for calling any API endpoint:

anysite api <endpoint> [key=value ...] [OPTIONS]

Parameters are passed as key=value pairs. Types are auto-converted using the schema cache.

# LinkedIn
anysite api /api/linkedin/user user=satyanadella
anysite api /api/linkedin/company company=anthropic
anysite api /api/linkedin/search/users title=CTO count=50 --format csv

# Instagram
anysite api /api/instagram/user user=cristiano
anysite api /api/instagram/user/posts user=nike count=20

# Twitter/X
anysite api /api/twitter/user user=elonmusk --format table

# Web parsing
anysite api /api/web/parse url=https://example.com

# Y Combinator
anysite api /api/yc/company company=anthropic

Output Formats

--format json    # Default: Pretty JSON
--format jsonl   # Newline-delimited JSON (for streaming)
--format csv     # CSV with headers
--format table   # Rich table for terminal

Field Selection

# Include specific fields (dot notation and wildcards supported)
anysite api /api/linkedin/user user=satyanadella --fields "name,headline,follower_count"

# Exclude fields
anysite api /api/linkedin/user user=satyanadella --exclude "certifications,recommendations"

# Compact JSON
anysite api /api/linkedin/user user=satyanadella --compact

Built-in field presets: minimal, contact, recruiting.

Save to File

anysite api /api/linkedin/search/users title=CTO count=100 --output ctos.json
anysite api /api/linkedin/search/users title=CTO count=100 --output ctos.csv --format csv

Pipe to jq

anysite api /api/linkedin/user user=satyanadella -q | jq '.follower_count'

Batch Processing

Process multiple inputs from a file or stdin:

# From a text file (one value per line)
anysite api /api/linkedin/user --from-file users.txt --input-key user

# From JSONL (one JSON object per line)
anysite api /api/linkedin/user --from-file users.jsonl

# From stdin
cat users.txt | anysite api /api/linkedin/user --stdin --input-key user

# Parallel execution
anysite api /api/linkedin/user --from-file users.txt --input-key user --parallel 5

# Rate limiting
anysite api /api/linkedin/user --from-file users.txt --input-key user --rate-limit "10/s"

# Error handling
anysite api /api/linkedin/user --from-file users.txt --input-key user --on-error skip

# Progress bar and stats
anysite api /api/linkedin/user --from-file users.txt --input-key user --progress --stats

Input file formats: plain text (one value per line), JSONL, CSV.

Dataset Pipelines

Collect multi-source datasets with dependency chains, store as Parquet, query with DuckDB, and load into a relational database. Includes per-source transforms, file/webhook exports, run history, scheduling, and webhook notifications.

Create a dataset

anysite dataset init my-dataset

Edit my-dataset/dataset.yaml to define sources:

name: my-dataset
sources:
  # Search sources (can be combined with union)
  - id: search_cto
    endpoint: /api/linkedin/search/users
    params: { keywords: "CTO fintech", count: 50 }

  - id: search_vp
    endpoint: /api/linkedin/search/users
    params: { keywords: "VP Engineering", count: 50 }

  # Union combines multiple sources (must have same endpoint)
  - id: all_candidates
    type: union
    sources: [search_cto, search_vp]
    dedupe_by: urn.value                  # Optional: remove duplicates by field

  # Dependent source using union as parent
  - id: profiles
    endpoint: /api/linkedin/user
    dependency:
      from_source: all_candidates
      field: urn.value
    input_key: user

  - id: companies
    endpoint: /api/linkedin/company
    from_file: companies.txt
    input_key: company
    transform:                          # Post-collection transform (for exports)
      filter: '.employee_count > 10'
      fields: [name, url, employee_count]
      add_columns:
        batch: "q1-2026"
    export:                             # Export to file/webhook after Parquet write
      - type: file
        path: ./output/companies-{{date}}.csv
        format: csv
    db_load:
      key: _input_value                    # Unique key for incremental sync
      sync: full                           # full (default) or append (no DELETE)
      fields: [name, url, employee_count]

  - id: employees
    endpoint: /api/linkedin/company/employees
    dependency:
      from_source: companies
      field: urn.value
    input_key: companies
    input_template:
      companies:
        - type: company
          value: "{value}"
      count: 5
    refresh: always                       # Re-collect every run with --incremental
    db_load:
      key: urn.value                       # Unique key for incremental sync
      sync: append                         # Keep old records (no DELETE on diff)
      fields: [name, url, headline]

storage:
  format: parquet
  path: ./data/

schedule:
  cron: "0 9 * * *"                    # Daily at 9 AM

notifications:
  on_complete:
    - url: "https://hooks.slack.com/xxx"
  on_failure:
    - url: "https://alerts.example.com/fail"

Collect, query, and load

# Preview collection plan
anysite dataset collect dataset.yaml --dry-run

# Collect data (supports --incremental to skip already-collected inputs)
anysite dataset collect dataset.yaml

# Collect and auto-load into PostgreSQL
anysite dataset collect dataset.yaml --load-db pg

# Check status
anysite dataset status dataset.yaml

# Query with SQL (DuckDB)
anysite dataset query dataset.yaml --sql "SELECT * FROM companies LIMIT 10"

# Query with dot-notation field extraction
anysite dataset query dataset.yaml --source profiles --fields "name, urn.value AS urn_id"

# Interactive SQL shell
anysite dataset query dataset.yaml --interactive

# Column stats and data profiling
anysite dataset stats dataset.yaml --source companies
anysite dataset profile dataset.yaml

# Load into PostgreSQL with automatic FK linking (incremental sync with db_load.key)
anysite dataset load-db dataset.yaml -c pg

# Drop and reload from latest snapshot
anysite dataset load-db dataset.yaml -c pg --drop-existing

# Load a specific snapshot date
anysite dataset load-db dataset.yaml -c pg --snapshot 2026-01-15

# Run history and logs
anysite dataset history my-dataset
anysite dataset logs my-dataset --run 42

# Generate cron/systemd schedule
anysite dataset schedule dataset.yaml --incremental --load-db pg

# Compare snapshots (diff two collection dates, supports dot-notation keys)
anysite dataset diff dataset.yaml --source employees --key _input_value
anysite dataset diff dataset.yaml --source profiles --key urn.value --fields "name,headline"

# Reset incremental state
anysite dataset reset-cursor dataset.yaml

Incremental Collection

When collecting data from from_file or dependency sources, anysite tracks which input values have already been processed. This allows resuming collection without re-fetching data you already have.

How it works:

1. After collecting a source, input values are saved to metadata.json (collected_inputs)
2. On next run with --incremental, these values are skipped
3. Only new input values are collected

# First run — collects all 1000 companies from file
anysite dataset collect dataset.yaml
# → Collected: 1000 records

# Add 50 new companies to the input file, run with --incremental
anysite dataset collect dataset.yaml --incremental
# → Skipped: 1000 (already collected), Collected: 50 (new only)

# Force re-collection of everything
anysite dataset reset-cursor dataset.yaml
anysite dataset collect dataset.yaml
# → Collected: 1050 records

Per-source control with refresh:

sources:
  - id: profiles
    refresh: auto      # (default) respects --incremental, skips collected inputs

  - id: activity
    refresh: always    # ignores --incremental, always re-collects
                       # useful for time-sensitive data (posts, activity feeds)

Reset cursor:

# Reset all sources — next run collects everything
anysite dataset reset-cursor dataset.yaml

# Reset specific source only
anysite dataset reset-cursor dataset.yaml --source profiles

Typical workflow for scheduled pipelines:

# Daily cron with incremental — only fetches new data
anysite dataset schedule dataset.yaml --incremental --load-db pg

# Weekly full refresh — reset and collect all
anysite dataset reset-cursor dataset.yaml && anysite dataset collect dataset.yaml --load-db pg

Database

Manage database connections and run queries.

# Add a connection (--password saves directly in connections.yaml)
anysite db add pg --type postgres --host localhost --database mydb --user app --password secret
# Or reference an existing env var
anysite db add pg --type postgres --host localhost --database mydb --user app --password-env PGPASS
# ClickHouse connection
anysite db add ch --type clickhouse --host ch.example.com --port 8443 --database analytics --user app --password secret --ssl
# Mark connection as read-only (prevents write operations)
anysite db add replica --type postgres --host replica.example.com --database mydb --user reader --read-only

# List and test connections
anysite db list
anysite db test pg

# Query
anysite db query pg --sql "SELECT * FROM companies" --format table

# Insert data (auto-create table from schema inference)
cat data.jsonl | anysite db insert pg --table users --stdin --auto-create

# Upsert with conflict handling
cat updates.jsonl | anysite db upsert pg --table users --conflict-columns id --stdin

# Inspect schema
anysite db schema pg --table users

Database Discovery

Introspect database schema, sample data, and optionally enrich with LLM descriptions:

# Discover schema (tables, columns, types, FKs, indexes, row counts, sample data)
anysite db discover mydb

# Discover with LLM-generated table/column descriptions and implicit relationship detection
anysite db discover mydb --with-llm

# Filter tables
anysite db discover mydb --tables users,posts --sample-rows 10
anysite db discover mydb --exclude-tables _migrations

# View saved catalogs
anysite db catalog                       # List all catalogs
anysite db catalog mydb                  # Show full catalog
anysite db catalog mydb --table users    # Show specific table
anysite db catalog mydb --json           # JSON output for agents

Read-only access is auto-detected during discovery. Use --read-only on db add to force it.

Supports SQLite, PostgreSQL, and ClickHouse. Passwords stored directly (--password) or via env var reference (--password-env).

LLM Analysis

LLM-powered analysis of collected dataset records. Summarize, classify, enrich, generate text, match records across sources, and find semantic duplicates.

pip install "anysite-cli[llm]"        # OpenAI + Anthropic SDKs

Setup

anysite llm setup

Configures provider (OpenAI or Anthropic), API key (paste directly or reference an env var), and default model. Tests the connection. Direct keys are saved in ~/.anysite/config.yaml.

Commands

# Classify records into categories (auto-detects categories if --categories omitted)
anysite llm classify dataset.yaml --source posts --categories "positive,negative,neutral" --format table

# Summarize each record
anysite llm summarize dataset.yaml --source profiles --fields "name,headline" --max-length 50

# Enrich records with LLM-extracted attributes
anysite llm enrich dataset.yaml --source posts \
  --add "sentiment:positive/negative/neutral" \
  --add "language:string" \
  --add "quality_score:1-10"

# Generate text using record fields as template variables
anysite llm generate dataset.yaml --source profiles \
  --prompt "Write a LinkedIn intro for {name} who works as {headline}" \
  --temperature 0.7

# Match records between two sources
anysite llm match dataset.yaml --source-a profiles --source-b companies --top-k 3

# Find semantic duplicates
anysite llm deduplicate dataset.yaml --source profiles --key name --threshold 0.8

Common options: --provider, --model, --fields, --format, --output, --parallel, --rate-limit, --temperature, --dry-run, --no-cache, --prompt, --prompt-file.

Cache

anysite llm cache-stats    # Show cache statistics
anysite llm cache-clear    # Clear all cached responses

Responses are cached in SQLite at ~/.anysite/llm_cache.db. Use --no-cache to skip cache lookup.

Configuration

Configuration is stored in ~/.anysite/config.yaml.

# Set a value
anysite config set api_key sk-xxxxx
anysite config set defaults.format table

# Get a value
anysite config get api_key

# List all settings
anysite config list

# Show config file path
anysite config path

# Initialize interactively
anysite config init

# Reset to defaults
anysite config reset --force

Configuration Priority

1. CLI arguments (--api-key)
2. Environment variables (ANYSITE_API_KEY)
3. Config file (~/.anysite/config.yaml)
4. Defaults

Global Options

anysite [OPTIONS] COMMAND

Options:
  --api-key TEXT       API key (or set ANYSITE_API_KEY)
  --base-url TEXT      API base URL
  --debug              Enable debug output
  --no-color           Disable colored output
  --json               Force JSON envelope output (auto-enabled in pipes)
  --human              Force human-readable output (override auto-JSON in pipes)
  --non-interactive    Disable interactive prompts (auto-enabled when stdin is not a TTY)
  --version, -v        Show version
  --help               Show help

Claude Code Skill

Install the anysite-cli skill for Claude Code to get AI-assisted data collection:

# Add marketplace
/plugin marketplace add https://github.com/anysiteio/agent-skills

# Install skill
/plugin install anysite-cli@anysite-skills

The skill gives Claude Code knowledge of all anysite commands, dataset pipeline configuration, and database operations.

Development

Setup

git clone https://github.com/anysiteio/anysite-cli.git
cd anysite-cli
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

# With dataset + database support
pip install -e ".[dev,data]"

Run Tests

pytest
pytest --cov=anysite --cov-report=term-missing

Linting

ruff check src/
ruff format src/
mypy src/

License

MIT