Mighty
Mighty · Coding
Mighty's Citadel is a server-side API that scans customer files and photos for AI-generated or edited fakes. Fraud and underwriting teams route each scan before a loan funds or a claim pays.
About Mighty
What Is Mighty
Mighty builds Citadel, an AI detection API that checks the files customers hand over during high-stakes workflows. It looks at paystubs, bank statements, damage photos, estimates, and invoices for signs of generated or edited content, and for hidden instructions that try to steer the next system. That is the job of an AI content authenticity API: catch a forged document before it reaches a decision. Mighty is the company and Citadel is the product, so you'll see both names side by side in the docs.
The problem Citadel targets is specific. Manual review misses digital forgery, and a single altered document can carry a loan from cash-flow analysis straight to funding. Citadel turns that check into a machine-readable signal that plugs into systems a team already runs. It doesn't replace a human decision.
One limit is worth stating up front. Citadel does not score whether a claim is true. It flags suspicious evidence and hidden instructions, and your team still decides what to do. If you want a truth oracle or a fact checker, this isn't it.
Getting Started
- Create an API key in the Mighty dashboard and name it by environment, like production claim intake or staging uploads.
- Keep the key on your server, then send a test file to the scan endpoint with a POST request to /v1/scan.
- Pick a scan phase, a mode, and a focus so Citadel knows what to look at and how deep to go.
- Read the action, risk score, and threats in the response, then route the file to continue, queue review, or stop.
- Book a working session with the team to scope commercial terms, since hosted plans are no longer sold self-serve.
Product Information
A quick look at Mighty's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Mortgage underwriting teams
- Insurance fraud investigators
- Claims adjusters
- Product and platform engineers
Tasks
- Verifying customer-submitted income documents
- Screening user uploads before storage
- Checking model or agent output
- Triaging a backlog of suspicious files
Scenarios
- A mortgage file arrives with a W-2 that looks fine until you zoom in
- A homeowner uploads storm-damage photos for a claim
- A jewelry appraisal accompanies a new insurance policy
- An AI agent reads an uploaded invoice
Key features
Single Scan Contract Across File Types
Citadel takes text, images, PDFs, documents, extracted text, model output, and agent tool output through one endpoint, so you don't stitch together separate detectors for each format. Audio transcripts work today, and audio file scanning is in closed beta. That means one integration covers a document queue and an agent's tool calls.
Allow, Review, or Block Routing
Every scan returns an action of ALLOW, WARN, or BLOCK, along with a risk score and a list of threats. Your app switches on the action to continue, queue review, or stop a workflow. The routing is deliberately simple so it drops into an existing decision path without a redesign.
Forensics Separate From Routing
Citadel keeps routing and forensics apart on purpose. A scan also returns an authenticity verdict of likely_real, likely_ai_generated, ai_generated, or indeterminate, plus a scan status of pending, complete, or failed. That separation matters because weak forensic evidence is still useful to a reviewer even when it isn't strong enough to block a file.
Hidden Instruction and Prompt Injection Detection
Files and images can carry low-contrast text or embedded instructions that aren't visible in normal rendering. Citadel flags these hidden instructions and prompt injection attempts, with a confidence score and an evidence excerpt for each one. For an agent that reads customer documents, that check is the difference between a tool call that follows your policy and one that leaks data.
Evidence and ID Chain for Audit
Scans return scan IDs, request IDs, scan group IDs, and session IDs. Group IDs connect related evidence, like an uploaded file and the text later extracted from it, while session IDs tie a wider loan, claim, or batch together. Reviewers and audit logs can then follow a file from intake to decision.
Modes and Focus Controls
You choose a mode to set scan depth and latency, and a focus that picks which image evidence to check. Focus options cover steganography-style hidden payloads, AI edits, and authenticity. The controls let a low-risk text message run cheap while a high-value photo gets a deeper pass.
Usage-Based Billing With a Free Preview
Citadel bills in Security Compute Units, where text, images, and PDF pages each draw a different amount. A free preview tier supports small tests, with a four-page PDF and one embedded image per request cap. Pro raises those ceilings to 1,000 pages and 100 embedded images. There's no public dollar rate, so real pricing is scoped on a call.
Pros and cons
Pros
- Runs server-side before funding, binding, or payout, so a flagged file never silently changes a decision.
- Returns a route plus attached evidence, which gives reviewers context instead of a bare yes or no.
- Handles text, images, PDFs, and agent output through one contract, cutting down on separate tools.
- Detects hidden instructions and prompt injection, a real risk once agents start reading customer files.
- Backs forged document detection and deepfake image detection on the same endpoint, so one fraud prevention API covers both.
- Ships client examples in curl, TypeScript, Python, and Go, so a first scan takes minutes.
Cons
- There's no self-serve sign-up or public price list, so you have to book a call before you can buy.
- It doesn't judge whether a claim or statement is true, so teams that expect a fraud verdict will be disappointed.
- Audio file scanning is still in closed beta, which leaves audio-heavy workflows partly uncovered.
- Billing in Security Compute Units takes some reading to predict, since image and PDF scans cost more than text.
Frequently asked questions
Citadel checks the file a customer submitted, such as paystubs, bank statements, applications, damage photos, estimates, and invoices, plus text already read from those files. It looks for generated or edited content and for hidden instructions that try to steer the next system, then returns allow, review, or block with evidence.
Related content
Explore related tools, skills, and articles for Mighty.
Mighty Alternatives
Forefront
Forefront · CodingForefront is a web platform for building with open-source AI. It lets you fine-tune leading open-source language models on your own data, evaluate how they perform, and run them through an API or export them to host yourself. Developers who want the convenience of a closed-source platform but insist on owning their models and data are the target audience here.
Startkit
StartKit.AI · CodingStartkit is a boilerplate for building AI SaaS and AI wrapper products. Think of it as an AI startup boilerplate with the boring parts already wired up: authentication, Stripe and Lemon Squeezy payments, usage limits, transactional email, and an AI API starter that talks to OpenAI, Anthropic, Groq, or Llama. You clone the repo, set your price, and start on the part of your product that people actually pay for. It's Next.js under React and Tailwind, so most of the boilerplate code already feels familiar.
Testim
Tricentis · CodingTestim is an AI-powered test automation platform for building and running end-to-end tests across web, mobile, and Salesforce applications. It leans on machine learning to keep tests stable when an interface changes, so teams spend less time fixing broken selectors. Not bad for an automated testing tool you can start using today. You create tests by recording actions in a browser, then optionally add JavaScript when you need more control. It's a solid pick for busy QA teams.