Skip to main content

Developers

Manipulation Detection API

A REST API that scores any text for 20 psychological manipulation tactics and returns the evidence behind every score — one POST, JSON in, JSON out.

20-category Influence Tactics Protocol scoring — the same engine behind Decipon's consumer tool, exposed as a REST API and an MCP server.

Request POST https://api.decipon.com/v1/analyze 
curl -X POST https://api.decipon.com/v1/analyze \
  -H "X-API-Key: sk_live_xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
  "content": "Officials warn the public must act immediately or face disaster."
}'
Response 200 OK · application/json 
{
  "hash": "a1b2c3d4e5f6",
  "overall_score": 47,
  "tier": "MODERATE",
  "confidence": 0.68,
  "top_factors": [
    {
      "category": "urgent_action_demands",
      "score": 4,
      "evidence": "must act immediately or face disaster"
    }
  ]
}

Authentication

Send your key in the X-API-Key header on every request. Keys are issued per credit purchase or B2B subscription — one key, metered usage, no OAuth dance.

Prefer bearer tokens? Authorization: Bearer <jwt> is also accepted. Every response echoes your balance in X-Credits-Remaining.

Reference

$0.03/call

Metered, billed monthly via Stripe. No minimum.

Get API access

Why this API

20 categories, not a sentiment score

Every response breaks down into the five composite factors and 20 categories of the Influence Tactics Protocol — emotional manipulation, suspicious timing, uniform messaging, tribal division, missing information.

Evidence per finding

Each category score comes with the spans of text that triggered it, so you can show users why — not just a number. Pair it with the validation account before you build on it.

Deterministic ML core

The 15 ML-scored categories are deterministic — same input, same score. The hybrid LLM tier (five categories) runs at temperature 0. Quick mode is ML-only; full mode adds LLM reasoning and alternative perspectives.

Endpoints

Full spec →
Method Path
POST /v1/analyze
POST /v1/analyze/url
POST /v1/analyze/batch
GET /v1/score/{hash}
GET /v1/consensus/{hash}
GET /v1/credits/balance
GET /health

Rate limits are per-IP and per-tier; quick analysis is capped higher than full analysis. Check your remaining credits any time with GET /v1/credits/balance or read the X-Credits-Remaining response header.

Model Context Protocol

Decipon also speaks MCP over streamable HTTP at /mcp — five tools (analyze_content, analyze_url, get_score, get_consensus, get_leaderboard) and two resources (decipon://methodology, decipon://categories). Agents can auto-discover it via /.well-known/mcp/server-card.json.

Same API key, same scoring engine — a projection of the REST API, not a reimplementation.

Publishers & badges

Embedding analysis on a news site? The Integration Guide covers badge embedding, domain verification, and the analytics dashboard. See publisher badges for the product overview.

Start a subscription

Ship it in an afternoon

One POST, JSON in, JSON out — $0.03 per call, metered, no minimum. Read the API reference, grab the OpenAPI spec, or get an API key.

Get API access