API Overview

FactFence is an evidence-cited hallucination scoring API for regulated AI outputs. Send any LLM response alongside its source document and get back a per-claim breakdown with verdicts (supported, contradicted, not_found, unverifiable), exact source-text evidence spans, and automatic correction suggestions for numeric/date mismatches.

No external AI calls — deterministic heuristic engine with trigram similarity scoring. Results are cached for 24 hours and fully auditable.

Tier Feature Matrix

Feature	Basic (Free)	Pro	Ultra	Mega
Single verify	✅	✅	✅	✅
Batch verify	—	Up to 10	Up to 25	Up to 25
`no_store` option	—	✅	✅	✅
Max claims / request	10	20	40	40
Caching (24 h)	✅	✅	✅	✅
Monthly verifications	50	5,000	50,000	Unlimited

Authentication

All endpoints except /v1/health require the X-RapidAPI-Key header, which is sent automatically when you subscribe. No additional setup needed.

Endpoints

GET /v1/health

Uptime check — no authentication required.

Request

GET /v1/health

Response

{
  "status": "ok",
  "version": "1.0.0",
  "engine": "heuristic-v1"
}

POST /v1/verify

Verify a single LLM response against a source document (text or URL). All tiers (Basic+).

Example — Basic tier with source_text

POST /v1/verify
Content-Type: application/json

{
  "response_text": "The company reported $4.2 billion in Q3 revenue.",
  "source_text": "In Q3 2025 the company reported revenue of $3.9 billion, up 12% year-over-year."
}

Example — Basic tier with source_url

POST /v1/verify
Content-Type: application/json

{
  "response_text": "The company reported $4.2 billion in Q3 revenue.",
  "source_url": "https://example.com/earnings-report.html"
}

Example — Pro+ tier with options

POST /v1/verify
Content-Type: application/json

{
  "response_text": "Revenue was $4.2 billion and net income rose 15%.",
  "source_text": "Revenue was $3.9 billion. Net income grew 12% to $840 million.",
  "options": {
    "no_store": true,
    "max_claims": 20
  }
}

Response

{
  "hallucination_score": 0.5,
  "coverage_score": 1.0,
  "claims": [
    {
      "index": 0,
      "text": "The company reported $4.2 billion in Q3 revenue.",
      "type": "numeric",
      "verdict": "contradicted",
      "evidence": [
        {
          "text": "In Q3 2025 the company reported revenue of $3.9 billion, up 12% year-over-year.",
          "score": 0.412,
          "offset": { "start": 0, "end": 80 }
        }
      ],
      "correction": {
        "original": "4.2",
        "corrected": "3.9",
        "source_quote": "In Q3 2025 the company reported revenue of $3.9 billion, up 12% year-over-year."
      }
    }
  ],
  "model": "heuristic-v1",
  "cached": false
}

POST /v1/batch/verify

Verify up to 25 items in a single request. Pro+ tiers only — Basic tier receives 403 Forbidden.

Tier	Max Items
Pro	10
Ultra	25
Mega	25

Request (Pro tier example)

POST /v1/batch/verify
Content-Type: application/json

{
  "items": [
    {
      "response_text": "The company reported $4.2 billion in Q3 revenue.",
      "source_text": "In Q3 2025 the company reported revenue of $3.9 billion."
    },
    {
      "response_text": "Unemployment fell to 3.2% in June.",
      "source_text": "The June jobs report showed unemployment at 3.5%, down from 3.7%."
    }
  ]
}

Response

{
  "results": [
    {
      "index": 0,
      "status": "ok",
      "result": {
        "hallucination_score": 0.5,
        "coverage_score": 1.0,
        "claims": [
          {
            "index": 0,
            "text": "The company reported $4.2 billion in Q3 revenue.",
            "type": "numeric",
            "verdict": "contradicted",
            "evidence": [
              {
                "text": "In Q3 2025 the company reported revenue of $3.9 billion.",
                "score": 0.45,
                "offset": { "start": 0, "end": 56 }
              }
            ],
            "correction": {
              "original": "4.2",
              "corrected": "3.9",
              "source_quote": "In Q3 2025 the company reported revenue of $3.9 billion."
            }
          }
        ],
        "model": "heuristic-v1",
        "cached": false
      }
    },
    {
      "index": 1,
      "status": "ok",
      "result": {
        "hallucination_score": 0.5,
        "coverage_score": 1.0,
        "claims": [
          {
            "index": 0,
            "text": "Unemployment fell to 3.2% in June.",
            "type": "numeric",
            "verdict": "contradicted",
            "evidence": [
              {
                "text": "The June jobs report showed unemployment at 3.5%, down from 3.7%.",
                "score": 0.38,
                "offset": { "start": 0, "end": 66 }
              }
            ],
            "correction": {
              "original": "3.2",
              "corrected": "3.5",
              "source_quote": "The June jobs report showed unemployment at 3.5%, down from 3.7%."
            }
          }
        ],
        "model": "heuristic-v1",
        "cached": false
      }
    }
  ],
  "total": 2,
  "completed": 2
}

GET /v1/models

List available verification engines. All tiers (Basic+).

Request

GET /v1/models

Response

{
  "models": [
    {
      "id": "heuristic-v1",
      "name": "Deterministic Heuristic Engine",
      "version": "1.0.0",
      "type": "heuristic",
      "description": "Trigram similarity + full-text search based verification. No external AI.",
      "roadmap": {
        "ml_reranker": "planned",
        "embedding_search": "planned"
      }
    }
  ]
}

GET /v1/pricing-hints

Returns suggested pricing tiers and limits. Useful for tier-aware client logic. All tiers (Basic+).

Request

GET /v1/pricing-hints

Response

{
  "plans": [
    { "name": "Basic",      "monthly_verifications": 50,     "price_usd": 0 },
    { "name": "Pro",        "monthly_verifications": 5000,   "price_usd": 29 },
    { "name": "Ultra",      "monthly_verifications": 50000,  "price_usd": 149 },
    { "name": "Mega",       "monthly_verifications": -1,     "price_usd": null, "contact": true }
  ]
}

Request / Response Schema

Verify Request Fields

Field	Type	Required	Description
`response_text`	string	Yes	The LLM output to verify
`source_text`	string	*	Ground-truth source document (provide this or `source_url`)
`source_url`	string	*	URL to fetch source text from (HTML stripped, 10 s timeout)
`options.no_store`	boolean	No	Skip caching & persistence (Pro+ only, default: false)
`options.max_claims`	integer	No	Max claims to extract — Basic: 10, Pro: 20, Ultra/Mega: 40
`options.chunk_size`	integer	No	Source chunk size in characters (default: 1000)
`options.chunk_overlap`	integer	No	Overlap between chunks (default: 150)

Verify Response Fields

Field	Type	Description
`hallucination_score`	number	0–1; higher = more hallucination detected
`coverage_score`	number	0–1; proportion of claims with source evidence
`claims[]`	array	Per-claim breakdown (see Claim Object below)
`model`	string	Engine used (e.g. `heuristic-v1`)
`cached`	boolean	Whether the result was served from cache

Claim Object

Field	Type	Description
`index`	integer	Claim position in response
`text`	string	Extracted claim sentence
`type`	string	`numeric` \| `date` \| `entity` \| `general`
`verdict`	string	`supported` \| `contradicted` \| `not_found` \| `unverifiable`
`evidence[]`	array	Matching source chunks with `text`, `score`, and `offset` (start/end)
`correction`	object\|null	If contradicted: `{ original, corrected, source_quote }`

Error Responses

Status	Meaning	Example
`400`	Bad Request — missing or invalid fields	`{"ok":false,"error":{"code":"VALIDATION_ERROR","message":"response_text is required"}}`
`401`	Unauthorized — invalid or missing API key	`{"ok":false,"error":{"code":"UNAUTHORIZED","message":"Invalid API key"}}`
`403`	Plan Required — feature not available on your tier	`{"ok":false,"error":{"code":"PLAN_REQUIRED","message":"Batch verify requires Pro or higher"}}`
`429`	Rate Limited — monthly quota exceeded	`{"ok":false,"error":{"code":"RATE_LIMITED","message":"Monthly verification limit reached"}}`

Pricing

Tier	Monthly Price	Verifications	Cost / Verify	Best For
Basic	$0	50	$0.00	Evaluation, prototyping
Pro	$29	5,000	~$0.006	Startups, internal tools
Ultra	$149	50,000	~$0.003	Production pipelines, compliance
Mega	$499	500,000	~$0.001	Regulated industries, on-prem

Practical Limits

Limit	Value	Notes
Source document size	250,000 characters	~62,500 words / ~125 pages
URL fetch timeout	10 seconds	HTML auto-stripped to plain text
Batch size	25 items (Pro: 10)	Max 50k total characters across all items
Batch timeout	25 seconds	Remaining items return `status: "timeout"`
Claims per request	10 / 20 / 40	Basic / Pro / Ultra+Mega
Cache TTL	24 hours	Bypass with `no_store: true` (Pro+)

Contact

📧 Email: factfence@fastmail.com
🔗 RapidAPI: rapidapi.com/bradsan42/api/fact-fence