Overview

The Context Windows API gives paid subscribers programmatic, read-only access to the same AI case study database that powers this site — search and filter 2,505 case studies (live count), pull full study details, and query proven impact aggregates.

All endpoints are HTTPS GET, return JSON, and live under a versioned base path:

https://www.contextwindows.ai/api/v1

API access is included in every paid plan — the same limits apply to all tiers. If you are on the free plan, see pricing.

Authentication

Create and manage API keys in Settings → API Keys. Keys look like cw_live_… and are shown exactly once at creation — store them securely. You can hold up to 5 active keys and revoke any of them at any time.

Pass the key on every request, either as a bearer token or an X-API-Key header:

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/case-studies?perPage=5"

# equivalent:
curl -H "X-API-Key: cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/case-studies?perPage=5"

Keys are validated against your subscription on every request: if your subscription lapses, your keys stop working immediately (and resume when it is active again).

Keep keys server-side. The API is designed for server-to-server use. Never embed a key in client-side JavaScript or a public repository — anyone who sees it can use your quota.

Rate limits

Limits are per key and identical on every plan:

Every successful response includes usage headers (X-RateLimit-Limit-Minute, X-RateLimit-Remaining-Minute, X-RateLimit-Limit-Day, X-RateLimit-Remaining-Day). Exceeding a limit returns 429 with a Retry-After header (seconds).

Responses & errors

Successful responses wrap the payload in data, with meta for pagination where relevant:

{
  "data": [ ... ],
  "meta": { "page": 1, "perPage": 25, "total": 3120, "totalPages": 125 }
}

Errors use one stable shape:

{
  "error": {
    "code": "invalid_request",
    "message": "Unknown parameter(s): industy. See /api-docs for supported parameters.",
    "details": { "unknownParams": ["industy"] }
  }
}

Understanding the data

The companies in a case study

Every case study describes an AI implementation, and up to three kinds of company can be involved in it. Case studies are structured from publicly published sources, so vendor and partner fields are only populated when the source identifies them.

Proven impact

The provenImpact score shows the quality of reported outcomes of a case study. It's a composite score where financial results (ROI, revenue, and cost savings) are weighted stronger than time savings and operational metrics: each proof type present contributes its weight (operational = 1, time_savings = 2, cost_soft = 4, cost_hard = 8, revenue = 16, ebit_roi = 32). A group's proven impact is the average of its studies' scores.

Two related fields appear alongside it in grouped responses: hardProofPercent — the share of a group's studies reporting hard financial outcomes (revenue or EBIT/ROI) — and proofTypeBreakdown — each proof type's contribution to the group's weighted score, as percentages summing to 100.

Agentic level

agenticLevel captures how autonomous the deployed AI is — from L1 (a tool people operate) to L5 (self-directed agent teams). null = not yet classified.

Freshness

New case studies are ingested daily. Dates are ISO-8601 strings in UTC.

Endpoints

List case studies

GET /api/v1/case-studies

Search, filter, sort, and paginate the case study database. All filters accept multiple values (repeat the parameter or comma-separate); multiple values within one filter are OR'd, different filters are AND'd.

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/case-studies?industry=Healthcare&sort=strongest_evidence&perPage=5"

{
  "data": [
    {
      "id": "a1b2c3…",
      "slug": "acme-health-fraud-detection",
      "url": "https://vendor.example.com/customers/acme-health",
      "publishedAt": "2026-03-14T00:00:00.000Z",
      "company": {
        "name": "Acme Health",
        "logo": "https://…",
        "website": "https://acmehealth.example.com",
        "country": "United States",
        "region": "North America",
        "industry": "Healthcare",
        "employeeRange": "5K-10K",
        "foundedYear": 1998
      },
      "useCase": { "name": "Fraud Detection", "slug": "fraud-detection", "group": "Analytics & Intelligence" },
      "businessFunction": "Finance",
      "businessValue": "Reduce Costs",
      "headline": "Acme Health cuts claim fraud losses 31% with AI screening",
      "summary": "Deployed an ML screening layer across 2.1M annual claims…",
      "headlineMetric": { "value": "31%", "modifier": "reduction in", "label": "fraud losses" },
      "resultHeadlines": ["31% reduction in fraud losses", "9x faster claim triage"],
      "proofTypes": ["cost_hard", "operational"],
      "provenImpact": 9,
      "agenticLevel": 3,
      "buildVsBuy": "buy",
      "hasPartner": false,
      "source": { "name": "VendorCo", "domain": "vendor.example.com" }
    }
  ],
  "meta": { "page": 1, "perPage": 5, "total": 214, "totalPages": 43 }
}

Query & pagination

Filters

Get a case study

GET /api/v1/case-studies/{slug}

The full study by its slug (from list responses): everything in the list shape plus the narrative (context / challenge / solution), all attributed quotes, all structured results, a richer company block, and the vendor and implementation partner blocks where identified (see Understanding the data).

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/case-studies/acme-health-fraud-detection"

{
  "data": {
    …everything from the list shape…,
    "company": { "id": "…", "name": "Acme Health", "description": "…", "location": "Boston, MA", … },
    "narrative": {
      "context": "Acme Health processes 2.1M claims a year…",
      "challenge": "Manual fraud review covered under 4% of claims…",
      "solution": "A screening model now scores every claim…"
    },
    "quotes": [
      { "text": "We recovered the project cost in 11 weeks.", "speakerName": "J. Doe", "speakerTitle": "VP Claims", "speakerCompany": "Acme Health" }
    ],
    "results": [
      { "headline": "31% reduction in fraud losses", "value": "31%", "modifier": "reduction in", "label": "fraud losses", "displayValue": "31%", "comparisonBaseline": null }
    ],
    "vendor": { "id": "…", "name": "VendorCo", "website": "https://vendor.example.com", … },
    "partners": [],
    "citations": null
  }
}

List use cases

GET /api/v1/use-cases

A use case is what the AI specifically does — the capability being deployed, like Fraud Detection or Document Drafting. It's independent of industry (the company's sector) and business function (the team using it): the same use case can show up in any industry or department. Every case study is mapped to one canonical use case.

This endpoint returns every use case with at least 5 case studies in the window, ranked by proven impact. Each row carries the use case's description, volume, proven impact, YoY growth, build-vs-buy split, agentic level, and the number of KPIs tracked across its studies. It takes its own parameters (below) — the case-study filters don't apply here.

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/use-cases?industry=Healthcare"

{
  "data": {
    "windowMonths": 24,
    "totalStudies": 812,
    "totalUseCases": 31,
    "useCases": [
      {
        "rank": 1,
        "name": "Clinical Documentation",
        "slug": "clinical-documentation",
        "group": "Document & Text",
        "description": "AI that drafts clinical notes and visit summaries from…",
        "volume": 74,
        "provenImpact": 9.8,
        "hardProofPercent": 18.92,
        "proofTypeBreakdown": { "operational": 9, "time_savings": 22, "cost_soft": 12, "cost_hard": 21, "revenue": 24, "ebit_roi": 12 },
        "growthPercent": 61,
        "buildVsBuy": { "classified": 58, "buildPercent": 22, "buyPercent": 78 },
        "agenticLevel": {
          "average": 2.3,
          "label": "L2",
          "classified": 66,
          "distribution": { "1": 14, "2": 29, "3": 18, "4": 5, "5": 0 }
        },
        "kpisTracked": 7
      }
    ]
  }
}

Notes: growthPercent is always last-12-months vs the 12 before, independent of months. buildVsBuy percentages are over classified studies only (null when none classified). kpisTracked is null when a use case doesn't yet have enough reported results. To pull the studies behind a row, pass the use case name to the list endpoint (?useCase=…).

Proven impact aggregates

GET /api/v1/aggregates/proven-impact

Compare proven impact, volume, and growth across any grouping: by use case, by industry, or by agentic level. Use this for cross-cutting comparisons — for the full per-use-case profile (description, build vs buy, agentic level, KPIs), use the use cases endpoint instead. Groups with fewer than 5 studies in the window are omitted (except agentic levels, which are always returned).

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/aggregates/proven-impact?group=useCase"

{
  "data": {
    "groupBy": "useCase",
    "windowMonths": 24,
    "totalStudies": 2960,
    "rows": [
      {
        "group": "Fraud Detection",
        "volume": 87,
        "provenImpact": 11.4,
        "hardProofPercent": 24.14,
        "proofTypeBreakdown": { "operational": 6, "time_savings": 9, "cost_soft": 12, "cost_hard": 27, "revenue": 31, "ebit_roi": 15 },
        "growthPercent": 42
      }
    ]
  }
}

Discover filter values

GET /api/v1/meta/filters

Live value lists (with document counts) for every filter dimension of the list endpoint — use it to populate dropdowns and validate inputs instead of hardcoding taxonomy values.

curl -H "Authorization: Bearer cw_live_YOUR_KEY" \
  "https://www.contextwindows.ai/api/v1/meta/filters"

{
  "data": {
    "industry": [ { "value": "Financial Services", "count": 412 }, … ],
    "useCase": [ { "value": "Support Chatbot", "count": 198 }, … ],
    "proofTypes": [ … ], "region": [ … ], "sort": [ … ], …
  }
}

OpenAPI spec

A machine-readable OpenAPI 3.1 description of the full API is available at:

https://www.contextwindows.ai/api/v1/openapi.json

Import it into Postman or Insomnia (File → Import → paste the URL) to get every endpoint, parameter, and response schema pre-configured, or feed it to an OpenAPI code generator for a typed client in your language.

Connect via MCP

Prefer to stay in your AI tools? Context Windows is also available as an MCP server — connect ChatGPT, Claude, Cursor, or any MCP-compatible assistant, and it can search case studies, pull full details, and rank use cases directly in your conversation, grounded in this data. There's nothing to install or download — it's a hosted server your tool connects to with one URL:

https://www.contextwindows.ai/mcp

No API key needed: when you add the connector, your tool opens a Context Windows sign-in and asks for your approval — access follows your paid subscription, with the same rate limits as the API (60 tool calls/min, 1,000/day). You can review and disconnect connected apps anytime in Settings → Connected apps.

Adding the connector

Available tools