Give your agent structured YouTube intelligence
Your content planning agent needs data it can act on, not raw view counts from the YouTube Data API. BrightBean returns scores, content gaps, and recommendations as structured JSON via REST API and MCP server.
The problem
Your agent is flying blind on YouTube
Scraping is brittle
Your agent scrapes YouTube, parses HTML, and guesses what's working. One layout change breaks everything. You spend more time maintaining scrapers than building features.
YouTube Data API = raw metrics
View counts and subscriber numbers tell you what happened. They don't tell you why it worked, what to do next, or whether a title is any good. Your agent can't make decisions from raw numbers alone.
With BrightBean
Structured JSON with scores, gaps, and rewrite suggestions. Your agent calls the API, gets intelligence, acts on it. No scraping, no guessing, no raw metric interpretation.
See it in action
Build a content planning agent in 20 minutes
Here's a typical workflow: your agent finds content gaps in a niche, scores title candidates, and builds a publishing calendar. All from structured API calls.
Connect via MCP or register as a tool
For MCP-compatible agents (Claude Desktop, Cursor, custom MCP clients), add one JSON config. Your agent auto-discovers all 5 BrightBean endpoints.
For LangChain, CrewAI, or OpenAI Agents, register BrightBean as a structured tool. The agent decides when to call each endpoint based on the user's request.
{
"mcpServers": {
"brightbean": {
"url": "https://api.brightbean.xyz/mcp",
"transport": "sse"
}
}
}from langchain.tools import StructuredTool
def score_title(title: str, niche: str) -> dict:
resp = requests.post(
"https://api.brightbean.xyz/v1/score/title",
headers={"Authorization": f"Bearer {API_KEY}"},
json={"title": title, "niche": niche}
)
return resp.json()
title_scorer = StructuredTool.from_function(
func=score_title,
name="youtube_title_scorer",
description="Score a YouTube title 0-100 for a niche"
)Agent discovers content opportunities
A user asks your agent: "Find me the best video topics in home cooking for next month." Your agent calls /content-gaps with the niche parameter.
The response includes topics ranked by demand score, competition count, and opportunity rating. Your agent can filter, sort, and present these in whatever format your user needs.
Each gap also includes a suggested title, pre-scored against the niche. Your agent has everything it needs to build a content calendar from a single API call.
{
"content_gaps": [
{
"topic": "5-ingredient slow cooker meals under $3",
"demand_score": 87,
"monthly_searches_est": 12400,
"competing_videos_above_10k_views": 3,
"opportunity_rating": "high",
"suggested_title": "5 Slow Cooker Meals Under $3 (Only 5 Ingredients)"
},
{
"topic": "meal prep for one person no waste",
"demand_score": 79,
"monthly_searches_est": 8900,
"competing_videos_above_10k_views": 5,
"opportunity_rating": "high",
"suggested_title": "Meal Prep for One: Zero Food Waste Method"
}
],
"niche": "home cooking",
"videos_analyzed": 14720
}Agent optimizes titles and builds the calendar
With content gaps identified, your agent takes the suggested titles and scores them. If a title scores below the niche average, it calls /score/title to get rewrite suggestions.
The agent can also call /benchmark to check how the user's channel compares to their niche, then factor that context into its recommendations.
End result: a complete publishing calendar with optimized titles, ranked by opportunity, personalized to the user's channel and niche. Built from 3-4 API calls.
What your agent outputs
Week 1: "5 Slow Cooker Meals Under $3"
Title score: 82 | Demand: 87 | Competition: Low
Week 2: "Meal Prep for One: Zero Food Waste"
Title score: 74 | Demand: 79 | Competition: Medium
Week 3: "I Made Every Meal From My Pantry for a Week"
Title score: 79 | Demand: 72 | Competition: Very Low
Architecture
Works in single-agent and multi-agent setups
Assign different BrightBean endpoints to different agents, or give one agent access to all of them. Each endpoint is independent and stateless.
Research agent
/content-gaps
Finds underserved topics in a niche. Passes results to the planning agent.
Optimization agent
/score/title + /score/thumbnail
Scores and rewrites titles. Checks thumbnails. Ensures content meets quality thresholds before publishing.
Analytics agent
/benchmark + /analyze/hook
Monitors channel performance against niche. Analyzes hook patterns for strategy adjustments.
Compatibility
Works with every major agent framework
/ Claude Code
Agents SDK
/ LangGraph
client
client
Title scoring
Under 500ms response time
Content gaps
1-3 seconds depending on niche size
Benchmarks
1-3 seconds per channel analysis
Common questions from agent developers
Can my LangChain agent call BrightBean directly? +
Yes. Register any BrightBean endpoint as a LangChain StructuredTool. Your agent decides when to call it based on the user's request. Works the same way with CrewAI, OpenAI Agents SDK, or any framework that supports HTTP tool calls.
What's the difference between MCP and REST API? +
MCP lets your agent discover and call BrightBean tools automatically with one config line. REST API gives you direct HTTP control for custom integrations. Same data, same endpoints. Choose whichever fits your architecture.
How fast are API responses? +
Title scoring returns in under 500ms. Content gaps and benchmarks take 1-3 seconds depending on niche size. Fast enough for real-time agent workflows where the user expects conversational response times.
Can I use BrightBean in a multi-agent system? +
Yes. Assign different endpoints to different agents: one for research (content gaps), one for optimization (title scoring), one for competitive analysis (benchmarks). Each agent makes independent API calls. All endpoints are stateless.
Does BrightBean replace the YouTube Data API? +
No. The YouTube Data API gives you raw data (views, likes, metadata). BrightBean gives you intelligence (scores, gaps, recommendations). Use both: YouTube API for the data, BrightBean for the decisions your agent needs to make.
What happens if my agent makes a lot of calls? +
Free tier: 500 calls. Standard: 100,000 calls/month at $99. Growth: 500,000 calls/month at $399. Overage is $0.02/call with no service interruption. For agent workloads, the Standard plan handles most production use cases.
Give your agent YouTube intelligence.
Free tier: 500 API calls, no credit card required.