DEV Community

TrustBoost-PII-Sanitizer
TrustBoost-PII-Sanitizer

Posted on

I Built a Competitive Intelligence API That AI Agents Pay For Autonomously — Here's How

#agents #ai #api #showdev

I previously built TrustBoost, a PII sanitization API for autonomous AI agent pipelines with x402 micropayments on Solana. Today I'm sharing Intelica — a competitive intelligence API that follows the same model. This is a technical walkthrough of what I built, why, and how you can use it or build something similar.

The Problem: Agents Are Flying Blind

Imagine you're building an AI agent that writes cold outreach emails. Your agent knows how to write. It knows the prospect's name and company. But before writing, it needs to understand what that company actually does, how they position themselves, who their competitors are, and what their users complain about.

Today most agents solve this one of two ways:

Option A — DIY analysis: The agent fetches the URL, strips the HTML, sends the raw text to an LLM, and tries to extract structured insights. This costs ~$0.08-0.15 in tokens, takes 8-12 seconds, and the developer had to build all that logic from scratch.

Option B — Skip it: The agent writes generic emails that don't convert because it has no real context.

There's a third option now.

The Solution: Pay-Per-Call Competitive Intelligence via x402

Intelica is a single HTTP endpoint that accepts any URL or text description of a product or company, runs a structured LLM analysis, and returns actionable competitive intelligence as JSON.

The twist: it uses the x402 protocol for payment. Your agent sends a request, receives an HTTP 402 with a USDC payment envelope, pays $0.05 autonomously, and immediately gets the analysis back. No accounts. No API keys. No subscriptions. No human intervention.

Here's what that looks like in practice:

# Step 1: Call the endpoint without payment
curl -X POST https://intelica.onrender.com/intel \
  -H "Content-Type: application/json" \
  -d '{"url": "https://figma.com"}'

# Returns HTTP 402:
{
  "x402Version": 1,
  "accepts": [{
    "scheme": "exact",
    "network": "base-mainnet",
    "maxAmountRequired": "50000",
    "payTo": "0x1d6bA7ac2461fd0E17D6A4C7bc1c9Ce365EfC4FF",
    "asset": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913"
  },
  {
    "scheme": "exact", 
    "network": "solana-mainnet",
    "maxAmountRequired": "50000",
    "payTo": "45q8KyCAGSHHd6qYP2ZkEJh22SzeQeXRfyFsENcx3KN6",
    "asset": "EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v"
  }],
  "error": "Payment required — pay with USDC on Base or Solana"
}

# Step 2: Agent signs payment, retries with X-PAYMENT header
# Step 3: Receive structured analysis
Enter fullscreen mode Exit fullscreen mode

The entire flow — request, 402, payment, verification, response — happens in under 3 seconds.

Real Output: What You Actually Get

Here's a live analysis of Figma I ran today:

{
  "company_or_product": "Figma",
  "positioning_summary": "Figma is a cloud-based collaborative design platform that enables product teams to design, prototype, and iterate together in real-time. It eliminates the need for design files to be passed between team members and provides a single source of truth for design work.",
  "target_customer": "Product design teams, UX/UI designers, product managers, and enterprises seeking collaborative design workflows",
  "core_value_props": [
    "Real-time collaborative editing and commenting",
    "Cloud-based accessibility across devices and locations",
    "Integrated prototyping and handoff capabilities",
    "Design system and component library management"
  ],
  "user_pain_points": [
    "Difficulty coordinating between distributed design teams",
    "Version control and file management complexity",
    "Inefficient designer-to-developer handoff processes",
    "Lack of centralized design system governance"
  ],
  "detected_competitors": [
    "Adobe XD", "Sketch", "Webflow", "Penpot", "InVision"
  ],
  "unique_angle": "Browser-based multiplayer editing that eliminates file versioning friction — a gap competitors haven't fully closed",
  "tone": "professional",
  "confidence": "high"
}
Enter fullscreen mode Exit fullscreen mode

Consistent structure. Every time. That's what makes it pipeline-friendly — your agent always knows exactly which fields to expect.

Why x402 Changes Everything for Agent Monetization

The x402 protocol revives the HTTP 402 status code — dormant since 1996 — and gives it a real job. Instead of returning an error, it returns a payment envelope describing exactly how much to pay, to which wallet, on which network.

The agent reads the envelope, signs the USDC transaction on Base or Solana, and retries the request with proof of payment. The server verifies on-chain via the Coinbase CDP Facilitator and returns the result.

Think of it like a vending machine:

  • The machine doesn't ask who you are
  • It doesn't make you create an account
  • You insert coins, it gives you the product
  • You walk away

Except the "coins" are USDC and the "machine" is an HTTP endpoint that any agent can call from anywhere in the world at any hour.

This is fundamentally different from traditional API monetization:

Traditional API x402 API
Requires registration Zero onboarding
Monthly subscription or prepaid credits Pay per call
Human sets up billing Agent pays autonomously
API key management No keys needed
Batch invoicing Instant USDC settlement

The Technical Stack

Intelica is intentionally minimal:

  • Runtime: FastAPI + Python 3.11
  • LLM: Claude Haiku — fast, cheap (~$0.0015/call), consistent JSON output
  • Payment: x402 protocol — USDC on Base mainnet + Solana mainnet
  • Cache: Supabase — results cached 6 hours, so repeated queries cost $0 in LLM
  • Audit log: Every paid call logged in Supabase with endpoint, source, cache hit, response time
  • Deploy: Render — auto-deploy from GitHub

The cache is the key efficiency lever. If 20 agents ask for a Notion analysis in the same 6-hour window, the LLM runs once and the other 19 get instant cached responses. Revenue: $1.00. LLM cost: $0.0015. Margin: 99.85%.

Three Endpoints, Three Use Cases

POST /demo — Free, 300 char limit, no URL fetching. Try it right now:

curl -X POST https://intelica.onrender.com/demo \
  -H "Content-Type: application/json" \
  -d '{"text": "Linear is a project management tool for software teams"}'
Enter fullscreen mode Exit fullscreen mode

POST /intel — $0.05 USDC via x402. Full analysis, URL support, Supabase audit log.

POST /batch — $0.20 USDC via x402. Up to 10 companies analyzed in parallel in a single call. For due diligence pipelines that need to process multiple competitors at once.

MCP Server for Claude Code Integration

Intelica also exposes a JSON-RPC 2.0 MCP server at /mcp with two tools:

{
  "method": "tools/call",
  "params": {
    "name": "analyze_competitor",
    "arguments": {
      "url": "https://notion.so",
      "context": "I'm building a note-taking app for developers"
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

Claude Code and any MCP-compatible agent framework can discover and call Intelica directly without knowing anything about x402 — the payment happens transparently at the transport layer.

Multi-Chain: Base and Solana

Intelica accepts payment on both Base mainnet and Solana mainnet. The 402 envelope lists both options — the agent chooses based on which network it operates on.

This matters because the two dominant ecosystems for agentic payments right now are Base (Coinbase AgentKit, x402 native) and Solana (Solana Agent Kit, Helius ecosystem). Covering both doubles the addressable market without any additional complexity for the caller.

Auto-Discovery: The /.well-known/x402.json Manifest

Intelica publishes a machine-readable manifest at:

https://intelica.onrender.com/.well-known/x402.json
Enter fullscreen mode Exit fullscreen mode

This is the file that x402 crawlers — x402scan, Agentic.Market, pay.sh — use to auto-index services without any manual registration. An agent searching for competitive intelligence tools will find Intelica through these directories without any human outreach required.

Who Actually Needs This

Five agent archetypes that would call Intelica today:

1. Outreach agents — Writing personalized cold emails requires understanding the prospect. An agent processing 50 prospects/day = 50 calls/day = $2.50/day from a single pipeline.

2. Due diligence agents — VCs and accelerators running automated startup evaluation. $0.05 per company is invisible against the value of the investment decision.

3. Content agents — SEO and content marketing agents that research the competitive landscape before writing. Understanding who else covers a topic and how they position is a prerequisite for differentiated content.

4. Product intelligence agents — Weekly competitive monitoring. 10 competitors analyzed via /batch = $0.20. One call, one payment, 10 structured reports.

5. Atelier agents — Intelica is listed on atelierai.xyz as a callable service. Other agents on the marketplace can subcontract competitive analysis to Intelica within their own workflows.

The Economics

Revenue per call: $0.05 USDC
LLM cost per call: $0.0015
Infrastructure: $7/month (Render starter)

Margin per call: 97%

At 1,000 calls/month: $50 revenue, $8.50 in costs, $41.50 net.
At 10,000 calls/month: $500 revenue, $22 in costs, $478 net.
At 50,000 calls/month: $2,500 revenue, $82 in costs, $2,418 net.

The margin improves with scale because the cache absorbs LLM costs. High-volume pipelines querying the same companies repeatedly pay full price but cost almost nothing to serve after the first hit.

Try It Now

Free demo — no payment, no setup:

curl -X POST https://intelica.onrender.com/demo \
  -H "Content-Type: application/json" \
  -d '{"text": "Your competitor description here"}'
Enter fullscreen mode Exit fullscreen mode

Full endpoint docs: https://intelica.onrender.com/docs

x402 manifest: https://intelica.onrender.com/.well-known/x402.json

Atelier listing: https://atelierai.xyz/services/intelica/competitive-intelligence-analysis-structured-report-in-secon

What's Next

The next features in the roadmap:

  • /track — Subscribe an agent to monitor a company for $0.50/month. When positioning changes are detected, a webhook fires automatically. Recurring revenue without recurring calls.
  • Reversible analysis — Store the raw source alongside the structured output so agents can re-analyze with different context without paying again.
  • SKILL.md — A Claude Code skill file that lets any Claude Code session discover and call Intelica as a built-in tool.

If you're building agents that need competitive context, or if you're building your own x402 service and want to discuss the implementation, drop a comment below.

Questions or feedback — drop them in the comments.

Built with FastAPI, Claude Haiku, Supabase, and the x402 protocol. Deployed on Render. Payments settle in USDC on Base and Solana.

Top comments (0)