Clarity

Company intelligence API for AI sales agents. One API call returns structured intelligence about any company: signals, contradictions, customer review sentiment, and evidence-backed outreach angles.

Live at clarityapi.co

What it does

Clarity takes a target company domain, researches it across 7 data sources in parallel, and returns:

• Company profile: industry, stage, and description
• Sales signals with implications (e.g., "Trustpilot 1.8/5 vs Gartner 4.5/5" → "enterprise buyers happy, consumers aren't")
• Contradiction detection: cross-references website claims against GitHub, news, jobs, community, and reviews
• Customer review sentiment: ratings from G2, Gartner, Trustpilot, Capterra, and Reddit via SerpAPI
• Tech stack extracted from GitHub repos
• Hiring patterns from external ATS platforms (Greenhouse, Lever, Ashby)
• Evidence-based relevance scoring: honest fit assessment that says "no angle identified" when the evidence doesn't support one
• Suggested outreach email that references specific findings, not generic bridges

Quick start

# Clone and set up
git clone https://github.com/deepgori/clarity.git
cd clarity
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Configure
cp .env.example .env
# Add your API keys to .env

# Run
python main.py
# Open http://localhost:8000

API

POST /api/company

Analyze a company and return structured intelligence.

curl -X POST https://clarityapi.co/api/company \
  -H "Content-Type: application/json" \
  -d '{
    "domain": "datadog.com",
    "seller_domain": "grafana.com"
  }'

Request fields:

Field	Required	Description
domain	Yes	Target company domain
seller_domain	No	Your company domain (for relevance scoring and outreach)
context	No	Extra context for the analysis

Response:

{
  "success": true,
  "intelligence": {
    "company_name": "Datadog",
    "domain": "datadog.com",
    "what_they_do": "Cloud monitoring and security platform...",
    "industry": "Cloud Infrastructure / DevOps",
    "stage": "Public",
    "signals": [
      {
        "signal": "Customer reviews show Gartner 4.5/5 but Trustpilot 1.8/5",
        "implication": "Enterprise buyers satisfied, consumer experience is poor",
        "confidence": 0.7
      }
    ],
    "contradictions": [],
    "tech_stack": ["Go", "Rust", "TypeScript"],
    "hiring_signals": ["Hiring data unavailable - company likely uses enterprise ATS platforms"],
    "sales_strategy": {
      "recommended_angle": "...",
      "conversation_starter": "...",
      "relevance_score": 0.8,
      "relevance_reasoning": "..."
    },
    "overall_confidence": 0.85
  },
  "suggested_email": "...",
  "processing_time_ms": 23500
}

POST /api/compare

Same as /api/company but also generates a generic email for side-by-side comparison.

Architecture

Request → Parallel fetch (7 sources) → AI synthesis → Post-processing gates → Response

Data sources (all fetched in parallel):

Source	Method	What it provides
Website	Jina Reader (trafilatura fallback)	Company claims, positioning, product info
News	Google News RSS (NewsAPI fallback)	Recent announcements, funding, partnerships
GitHub	GitHub API	Repo activity, tech stack, commit freshness
Careers	Direct scraping	Internal job listings, team structure
External Jobs	Greenhouse, Lever, Ashby APIs	Department breakdown, tech requirements
Community	Hacker News Algolia API	Developer sentiment, public criticism
Reviews	SerpAPI (Google search)	G2, Gartner, Trustpilot, Capterra ratings; Reddit threads

AI layer:

• GPT-4o with structured JSON output for intelligence synthesis
• GPT-4o for evidence-driven email generation
• Cross-source contradiction detection
• Evidence-based relevance scoring (honest "no angle" when fit is weak)

Post-processing gates (deterministic, code-level):

• Enterprise ATS filter: suppresses false contradictions from missing job data for public companies
• Hiring inference guard: prevents fabricated claims about job postings when no data exists
• Banned phrase scrubber: removes AI-isms from output text

Configuration

Variable	Required	Description
OPENAI_API_KEY	Yes	OpenAI API key
SERPAPI_KEY	Yes	SerpAPI key for customer review data
CLARITY_API_KEY	No	API key for authentication (disabled if not set)
CLARITY_GITHUB_TOKEN	No	GitHub token for higher rate limits
NEWS_API_KEY	No	NewsAPI key for news fallback

Cost per query

Component	Cost
SerpAPI (reviews)	~$0.01
OpenAI GPT-4o (synthesis + email)	~$0.03-0.05
All other sources	Free
Total	~$0.04-0.06

License

MIT