WANT2VIEW CLI + MCP

One command gives Codex, Claude, Cursor, OpenClaw, and terminal agents the same WANT2VIEW evidence your account can access. MCP mode gives Codex live WANT2VIEW tools for YouTube, TikTok, and Telegram research.

WANT2VIEW CLI is the open-source agent interface for your WANT2VIEW account. It turns keyword searches, channel audits, local CSV/JSON imports, public catalog samples, and WANT2VIEW projects into source-of-truth packs your agents can read directly.

Start from your WANT2VIEW account in 60 seconds. CLI and MCP do not create a separate pricing model: they are the agent interface to the same dashboard limits, sources, projects, and API access.

npx -y want2view connect codex --token w2v_... --remote --skip-research

That single npm command installs the Codex skill, saves the token privately, and connects Codex to the remote WANT2VIEW MCP server. After that, tell Codex your niche in plain language.

Use WANT2VIEW MCP. My niche is AI SaaS/avatar/content-factory videos.
Find what videos I should make next: hooks, themes, visuals, and scripts with evidence URLs.

Pick any agent:

npx want2view claude https://youtube.com/@example --channel
npx want2view codex "b2b saas launch" # Cursor
npx want2view codex "ugc ads" # OpenClaw

Open-source agent connector. Same WANT2VIEW account, same product limits.
Website · Create account · Get API access

Workflows · Example Outputs · Launch Posts · Launch Checklist · API Access Flow · Launch Playbook · Support

Setup In 60 Seconds

For Codex, copy the one-command setup from WANT2VIEW API Access:

npx -y want2view connect codex --token w2v_... --remote --skip-research

Then restart Codex and ask it to use WANT2VIEW MCP. Local stdio mode stays available when you need exact control.

Agent	One command
Codex remote MCP setup	`npx -y want2view connect codex --token w2v_... --remote --skip-research`
Codex remote MCP URL	`codex mcp add want2view --url https://api.want2view.com/api/v1/developer/mcp`
Codex local MCP mode	`codex mcp add want2view -- npx -y want2view mcp`
Codex setup	`npx -y want2view install codex`
Codex	`npx want2view codex "ai video ads"`
Claude	`npx want2view claude "ugc ads"`
Cursor	`npx want2view codex "b2b saas launch"`
OpenClaw	`npx want2view codex "ugc ads"`
Claude channel audit	`npx want2view claude https://youtube.com/@example --channel`
Any terminal agent	`npx want2view codex "your keyword"`

Run account research from the same WANT2VIEW plan:

npx -y want2view account research "AI SaaS avatar content factory" --sources youtube,tiktok,telegram --limit 30 --wait 180

The command creates a local .want2view/exports/<pack_id>/ folder and prints the exact prompt to paste into your agent.

For a production decision, use the corpus mode. It runs independent live queries, deduplicates canonical URLs, and makes the data gaps explicit. It uses the same account limits; it does not create a CLI-specific plan.

npx -y want2view account research "what AI SaaS videos should we make next" \
  --sources youtube,tiktok,telegram --production --limit 80 --per-query-limit 60

The resulting pack adds production_coverage.json and production_queue.json: raw/unique counts, source coverage, transcript readiness, and the 20-80 URLs that Codex must open before making screen-level claims. Catalog fallback is off by default and can only be enabled explicitly with --allow-catalog-fallback.

MCP mode exposes WANT2VIEW as live tools:

doctor checks auth and connector state.
create_research_plan turns a business goal into short source queries, waits, exports, and merges one evidence pack.
create_production_corpus in local MCP builds a live deduplicated corpus and a subtitle/visual-review queue for production decisions.
export_research_plan in remote MCP merges completed plan runs into the same honest corpus after Codex has checked their status.
create_research starts a focused WANT2VIEW account research run.
get_status checks source status and record counts.
export_pack downloads an agent-ready evidence pack.
get_subtitles queues/reads the subtitles + scenario pipeline for selected videos.
search_telegram starts a Telegram-only research run and reports plan/source limits honestly.

Why do agents have different commands? They all read the same WANT2VIEW evidence folder. The command only chooses the most useful handoff file:

Codex, Cursor, OpenClaw, and generic terminal agents can use the Codex-style task pack: codex_tasks.md.
Claude gets claude_brief.md.
Every pack still includes manifest.json, agent_contract.md, status.md, summary.md, evidence.jsonl, and scored.csv.

Same Prompt. Different Intelligence.

Without WANT2VIEW CLI	With WANT2VIEW CLI
Agent guesses from a vague prompt	Agent reads `evidence.jsonl`, `scored.csv`, and `summary.md`
No source rows	Every recommendation can cite evidence
Manual copy-paste research	One local pack for Codex, Claude, Cursor, OpenClaw, and other terminal agents
Separate tool limits	One WANT2VIEW account plan across dashboard, CLI, MCP, and API

TL;DR

WANT2VIEW CLI is not a separate product tier. It is the open-source agent interface for your WANT2VIEW account.

Free WANT2VIEW account: create a Developer token, connect Codex/Claude, run test-depth YouTube/TikTok/Telegram research, and export evidence packs.
Paid WANT2VIEW account: the same CLI/MCP commands run with higher record limits, saved projects, scheduled refreshes, project API, team workflows, bulk reports, and custom research.
Best use: give Codex, Claude, and terminal agents real WANT2VIEW evidence folders instead of vague prompts.

What The Command Creates

npx want2view codex "ai video ads"

Output:

.want2view/exports/<pack_id>/
  manifest.json
  agent_contract.md
  status.md
  summary.md
  evidence.jsonl
  scored.csv
  codex_tasks.md
  claude_brief.md
  upgrade_prompt.md
  production_coverage.json  # production corpus mode
  production_queue.json     # selected URLs for subtitles + screen review

Then tell your agent:

Use the newest .want2view/exports/<pack_id> folder as the source of truth.
Read manifest.json, summary.md, evidence.jsonl, scored.csv, and codex_tasks.md.
Base recommendations only on evidence rows.

Want to inspect the output before running anything? Open examples/outputs.

Prompt Library

Copy these after running start.

Goal	Prompt
Codex strategy	`Use the newest .want2view/exports folder as the source of truth. Read manifest.json, summary.md, evidence.jsonl, scored.csv, and codex_tasks.md. Recommend only from evidence rows.`
Claude brief	`Read claude_brief.md and evidence.jsonl. Create a decision-ready content brief with hooks, risks, gaps, and next experiments. Mark assumptions clearly.`
Landing page	`Use the WANT2VIEW pack to write landing copy. Tie claims to evidence rows and separate observed patterns from creative recommendations.`
Content scripts	`Create 10 short-form scripts from the strongest evidence. For each script, cite the row or URL that inspired it.`

Pick Your Path

You want to...	Run this	What happens
Install Codex workflow	`install codex`	Adds the WANT2VIEW research skill to local Codex
Connect Codex live tools	`mcp` through `codex mcp add`	Lets Codex call WANT2VIEW research/status/export/subtitle/Telegram tools
Start with Codex in one command	`start codex "ai video ads"`	Creates a pack and prints the Codex prompt
Start with Claude in one command	`start claude https://youtube.com/@example --channel`	Creates a channel pack and prints the Claude prompt
Try without a token	`search "ai video ads" --demo`	Creates a local sample pack so an agent can inspect the file format
Try a channel-shaped pack	`channel https://youtube.com/@example --demo`	Creates a local channel-shaped sample pack
Use your own data	`import ./competitors.csv`	Converts CSV/JSON/JSONL into an evidence pack
Give Codex better context	`export --for codex`	Writes `codex_tasks.md`, `summary.md`, `evidence.jsonl`
Give Claude a strategy brief	`export --for claude`	Writes `claude_brief.md` with evidence references
Browse the public catalog	`catalog categories`	Shows a limited public sample of WANT2VIEW categories
Export public catalog evidence	`catalog export ai --for codex`	Gives an agent-readable sample pack from public catalog data
Use your own WANT2VIEW projects	`projects list`	Requires `WANT2VIEW_PUBLIC_API_KEY` from API Access
Export a project to an agent	`project export <id> --for codex`	Turns your internal project into an AI-readable pack
Run account research	`login` then `account research "goal"`	Splits broad goals into short queries, waits, exports, and merges one pack inside your account limits
Order custom research	Open WANT2VIEW	Get a human/product-ready category research package
Pick a ready workflow	`workflows keyword`	Shows commands, agent prompt, and WANT2VIEW next step

One Account Plan

Surface	Free WANT2VIEW account	Paid WANT2VIEW account
Dashboard	Free catalog, onboarding, API Access, Developer token	Saved projects, reports, history, team workflows
CLI	Local packs, CSV/JSON imports, test-depth account research	Higher record limits and project exports through the same commands
MCP	Codex can call WANT2VIEW research/status/export/subtitle/Telegram tools	Codex can run deeper repeated research inside paid limits
Sources	YouTube, TikTok, Telegram test depth	Deeper YouTube/TikTok/Telegram collection and roadmap sources when enabled
Upgrade reason	When the first pack proves value	More records, scheduled refresh, project API, bulk reports, team access

Why This Exists

Most AI agents are only as good as the context you give them. WANT2VIEW CLI turns social content evidence into clean, inspectable files that agents can use as a source of truth.

Without an account token, the CLI still gives immediate local value for format testing:

sample research packs;
CSV / JSON / JSONL imports;
basic normalization and scoring;
Codex-ready task packs;
Claude-ready research briefs.

Your WANT2VIEW account plan controls the connected product:

test-depth YouTube/TikTok/Telegram research through Developer tokens on free accounts;
deeper catalog data and higher record limits on paid plans;
private project exports from the WANT2VIEW dashboard;
managed YouTube, TikTok, and Telegram connectors;
roadmap source expansion only when enabled by plan/API;
provider keys, retries, source warnings, and cost controls;
scheduled refreshes;
team workspaces;
historical source index;
hosted dashboard and API;
custom research packages for teams that want the result done for them.

One-Command Agent Interface

Npm command:

npx want2view codex "ai video ads"
npx want2view claude https://youtube.com/@example --channel
npx want2view workflows keyword

NPM is the stable path:

npx -y want2view connect codex --token w2v_... --remote --skip-research

Use GitHub only when testing unreleased connector fixes.

The fastest real path is account-first: open WANT2VIEW API Access, create a Developer token, connect MCP once, then let Codex run inside your WANT2VIEW limits. Local sample packs are only there so an agent can inspect the format before login.

Agent Handoff

Tell Codex, Claude Code, or another terminal agent:

Use the newest .want2view/exports/<pack_id> folder as the source of truth.
Read manifest.json, summary.md, evidence.jsonl, scored.csv, and codex_tasks.md.
Base recommendations only on evidence rows.

Generated packs contain:

manifest.json
summary.md
evidence.jsonl
scored.csv
codex_tasks.md
claude_brief.md

More copy-paste prompts live in docs/AGENT_PROMPTS.md.

Workflows

Workflows are not separate hidden features. They are practical paths that combine the same WANT2VIEW product surfaces: keyword search, channel analysis, project export, catalog access, and account connectors.

Workflow	Start with CLI	Continue in WANT2VIEW
Keyword search to Codex	`npx want2view codex "ai video ads"`	Connect API Access when you want live account data
Channel audit to Claude	`npx want2view claude https://youtube.com/@example --channel`	Connect your account for real collection or order a deeper audit
Keyword research to Cursor	`npx want2view codex "creator economy"`	Turn evidence into product, SEO, or content tasks
Research pack to OpenClaw	`npx want2view codex "ugc ads"`	Work from the same source-of-truth files
WANT2VIEW project export	`projects list` then `project export <id> --for codex`	Work from private projects and team workspaces
Content team monitoring	`account research "category"`	Run inside account limits; paid plans add refreshes and team packs

Full copy-paste workflows live in docs/RECIPES.md.

Interactive Login

There are two keys because there are two product surfaces:

WANT2VIEW_API_TOKEN: Developer CLI token for account research runs.
WANT2VIEW_PUBLIC_API_KEY: Public API key for your WANT2VIEW projects and reports.

Account research needs a WANT2VIEW Developer CLI token. The login wizard gives users three paths:

npx want2view login

Choose authentication method:
  1. Open browser and create/paste Developer CLI token
  2. Paste API token now
  3. Use WANT2VIEW_API_TOKEN from environment
  4. Skip for now

Fast paths:

npx want2view login --token w2v_your_token
export WANT2VIEW_API_TOKEN="w2v_your_token"

Credential safety: token-bearing commands ignore project-level API base overrides. A cloned repository cannot redirect your WANT2VIEW_API_TOKEN or WANT2VIEW_PUBLIC_API_KEY to an arbitrary host through local config. Non-WANT2VIEW API endpoints require an explicit --allow-untrusted-api flag.

Agent diagnostics:

npx want2view doctor --json

Project API:

export WANT2VIEW_PUBLIC_API_KEY="your_dashboard_api_key"
npx want2view projects list
npx want2view project export <project_id> --for codex

WANT2VIEW Catalog

Use catalog commands to discover a limited public sample of what WANT2VIEW already tracks:

npx want2view catalog categories
npx want2view catalog videos ai --limit 20
npx want2view catalog export ai --for claude

This is the bridge from GitHub discovery into the real WANT2VIEW product: the CLI can show public catalog surfaces, while authenticated users can export private projects, run account research, and unlock deeper social intelligence.

Account Research

Once authenticated, the CLI can ask your WANT2VIEW account to run managed social collection. Use this for broad Codex/Claude tasks:

npx want2view account research "fitness reels content strategy" \
  --sources youtube,tiktok,telegram --limit 30 --wait 180

The account command creates a short query plan, starts several focused WANT2VIEW runs, waits for source jobs, exports the finished runs, and merges one evidence pack for the agent.

Low-level run commands remain available when you already know the exact keyword:

npx want2view cloud status w2v_run_abc123
npx want2view cloud export w2v_run_abc123 --for codex

The CLI stores exported packs locally under:

.want2view/exports/<run_id>/

Account Capabilities

The CLI/MCP connector follows your WANT2VIEW account plan:

Plan capability	Best for
Account connector	First Codex research run, up to the current account record limit, using YouTube/TikTok/Telegram
Project API	Agencies, AI agents, automations, and internal tools that need project data when included by the account plan
Managed connectors	Teams that need deeper YouTube, TikTok, and Telegram without maintaining provider access
Scheduled refreshes	Weekly monitoring of keywords, channels, creator markets, product categories, or competitor sets
Custom research	Done-for-you category reports, competitor maps, content briefs, scripts, and client-ready insights
Team workspace	Shared source-of-truth packs, saved history, and repeatable research workflows

Start here:

Website: want2view.com
App / API token: app.want2view.com
API Access: app.want2view.com/api-access
Custom research: open a WANT2VIEW account and request a research package from the dashboard.

Built For These Workflows

Content teams researching a new category before filming.
Agencies turning competitor data into client-ready strategy.
Founders and operators checking keywords, channels, and categories before spending budget.
AI agents that need evidence files before writing briefs, scripts, landing pages, or reports.
Internal tools that need a clean bridge into WANT2VIEW data.

Bring Your Own Data

npx want2view import examples/competitors.csv
npx want2view score
npx want2view export --for claude

Supported local inputs:

.csv
.json
.jsonl

Recommended columns:

Column	Meaning
`platform`	YouTube, TikTok, Telegram, or custom source
`account`	Creator, channel, author, subreddit, or profile
`title`	Content title or post headline
`url`	Source URL
`views`	Reach metric
`likes`	Likes or equivalent reaction count
`comments`	Comment count
`published_at`	Publication date
`text`	Caption, summary, or note

Command Map

Command	Purpose
`want2view start codex "topic"`	One-command Codex setup
`want2view start claude <channel> --channel`	One-command Claude channel setup
`want2view login`	Interactive auth wizard
`want2view doctor --json`	Agent-readable setup diagnostics
`want2view workflows`	Show keyword, channel, project, and monitoring workflows
`want2view workflows keyword`	Print a copy-pasteable keyword-to-agent workflow
`want2view search "topic" --demo`	Local sample keyword pack
`want2view channel <url> --demo`	Local sample channel-shaped pack
`want2view catalog categories`	Browse WANT2VIEW catalog categories
`want2view catalog export ai --for codex`	Export catalog evidence for an agent
`want2view projects list`	List your WANT2VIEW projects with `WANT2VIEW_PUBLIC_API_KEY`
`want2view project export <id> --for codex`	Export your internal project to Codex/Claude
`want2view install codex`	Install the WANT2VIEW research skill into local Codex
`want2view account research "goal"`	Plan, run, wait, export, and merge account research for an agent
`want2view account status <run_id>`	Poll an underlying account research run
`want2view account export <run_id> --for codex`	Download an underlying account context pack
`want2view research "topic" --demo`	Backward-compatible alias for local demo research
`want2view import ./file.csv`	Bring your own data
`want2view score`	Score local records
`want2view export --for codex`	Create Codex pack
`want2view export --for claude`	Create Claude brief
`want2view cloud research "topic" --mode cloud`	Low-level focused run used by `account research`
`want2view cloud status <run_id>`	Backward-compatible status alias
`want2view cloud export <run_id> --for codex`	Backward-compatible export alias

Account Plan Boundary

Open-source:

CLI runner;
local imports;
sample data;
limited public catalog access;
normalization;
basic scoring;
Codex and Claude context-pack exports;
auth wizard and account API client.

WANT2VIEW account plans:

Free account: Codex connector, Developer token, test-depth YouTube/TikTok/Telegram research, evidence packs.
Paid account: higher limits, saved/private projects, project API, scheduled refreshes, bulk reports, team workspaces, deeper scoring, historical indexes.

Local Development

node bin/want2view.js start codex "ai video ads"
node bin/want2view.js doctor --json

Name		Name	Last commit message	Last commit date
.github		.github
assets		assets
bin		bin
docs		docs
examples		examples
.gitignore		.gitignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
ROADMAP.md		ROADMAP.md
SECURITY.md		SECURITY.md
SUPPORT.md		SUPPORT.md
package-lock.json		package-lock.json
package.json		package.json

Navigation Menu

Repository navigation

About

Repository files navigation

WANT2VIEW CLI + MCP

Setup In 60 Seconds

Same Prompt. Different Intelligence.

TL;DR

What The Command Creates

Prompt Library

Pick Your Path

One Account Plan

Why This Exists

One-Command Agent Interface

Agent Handoff

Workflows

Interactive Login

WANT2VIEW Catalog

Account Research

Account Capabilities

Built For These Workflows

Bring Your Own Data

Command Map

Account Plan Boundary

Local Development

Links

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

WANT2VIEW CLI + MCP

Setup In 60 Seconds

Same Prompt. Different Intelligence.

TL;DR

What The Command Creates

Prompt Library

Pick Your Path

One Account Plan

Why This Exists

One-Command Agent Interface

Agent Handoff

Workflows

Interactive Login

WANT2VIEW Catalog

Account Research

Account Capabilities

Built For These Workflows

Bring Your Own Data

Command Map

Account Plan Boundary

Local Development

Links

About

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages