Skip to main content
PHOTOTOLOGY
PERSISTENT VISUAL INTELLIGENCE

The proof ıs ın the pıcture.

The memory layer for image intelligence.

Photo in. Structured intelligence out. Sealed to the registry. Free lookups forever.

Agent fuel.

Drop a photo. Right here.
No signup. No credit card. Results in seconds.

Built for agent memory.

We're everything AI isn't.

Immutable
Sealed on write. Your agent pipeline won't break because a finding changed.
Deterministic
Same photo, same answer. Even when underlying models update. Your pipelines stay stable.
Repeatable
Every lookup after the first is free. Zero tokens. Zero latency.
Idempotent
Retry as many times as you need. No side effects. Safe for autonomous agents.
Accretive
Only adds, never subtracts. Add a new lens later without re-running the old ones.
Provenant
Every finding traced to its evidence. Which lens found it. What supported it. Your agent can reason about confidence.
Attestable
Cryptographically signed. Any downstream system can verify the analysis wasn't tampered with.

Analyze. Structure. Remember.

Analyze

16 lenses extract structured intelligence from the photo.

Structure

Every finding validated against typed schemas. Sealed and signed.

Remember

Indexed by perceptual hash. Cached forever. Embedded in the photo's EXIF.

One engine. Every domain.

The Wikipedia of images.

Damaged roof with missing shingles and structural breach analyzed by Phototology for insurance claim
Describe (insurance)ConditionEntitiesQuality
Photo in. Claim out.

Catastrophic structural breach in residential roof. Multiple missing shingle sections with exposed decking. Wind and impact damage consistent with severe weather event.

SeveritySevere
Damage typeWind + impact
Affected areaRoof (northwest section)
Structure exposedYes (decking visible)
Recommended actionFull roof replacement
Claim priorityHigh

How your agent uses it.

Query the registry first. Analyze only on miss. Every analysis becomes free cache for the next agent.

Agent encounters a photo
GET /v1/lookup?hash=sha256(image)
Cache hit
Instant. $0.00. Done.
Cache miss
POST /v1/analyze with lenses.
$0.01-$0.20 per photo.
Stored in registry. Next agent gets it free.
import { Phototology } from '@phototology/sdk';
const ph = new Phototology({ apiKey });

// Check the registry first
const cached = await ph.lookup(photo);
if (cached) return cached;  // free, instant

// Analyze and store on miss
return await ph.analyze(photo, {
  lenses: ['describe', 'dating', 'entities']
});

A penny per lens.

Pick the modules you need. Each one costs $0.01. Full analysis runs under $0.20 per photo.

1,000 free calls per month. No credit card required.

Lookups from the registry are always free.

See the docs

Works with

  • Claude CodeMCP
  • Claude DesktopMCP
  • CursorMCP
  • VS CodeMCP
  • AntigravityMCP
  • CodexMCP

Ship in minutes.

Sandbox key included. No signup required.

GET /v1/lookup?hash=abc

Analyze once. Look up forever. Free.

MCP Server

Copied!

Drop in. Works immediately.

TypeScript SDK

Copied!

Typed client. First call in 3 lines.

REST API

Copied!

Sandbox key included. No signup.

Web App

analyze.phototology.com →

Upload a photo. See results now.

Powering 3 products in production

Dr. Vin

AI vehicle inspection reports

Candela

Memorial storytelling

Phossil

Family photo archive

Frequently Asked Questions

What happens to my photos?

Your photos are processed, enriched with metadata, and returned to you. Phototology does not store full-resolution images beyond 90 days. Analysis results and your photo registry are kept permanently. You can delete your photos and all associated data at any time.

How accurate is the date estimation?

Phototology identifies 23 categories of date anchors: vehicles, clothing, technology, signage, photographic processes, and more. Every estimate comes with evidence chains showing exactly what signals informed the conclusion. Add more modules and the estimate sharpens. A single module might say '1990s.' Add Entities and Location, and a Honda CR-V narrows it to 1997-2001. Add Text, and a concert poster locks it to Fall 1997.

Is my data used to train AI models?

No. Phototology does not use your photos to train, fine-tune, or improve any AI model. Your photos are analyzed using third-party AI providers under their paid API tiers, which contractually prohibit training on customer data.

How does pricing work?

$0.01 per lens per image. Every surface: web UI, API, SDK, MCP server. 16 lenses available. Full analysis under $0.20 per photo. 1,000 free credits per month, no credit card required.

How is this different from Google Cloud Vision?

Traditional photo AI (Google Cloud Vision, AWS Rekognition) returns labels: 'person (0.97)', 'outdoors (0.94)'. Phototology returns structured context: date estimates with evidence chains, domain-specific narratives, condition assessments, and metadata embedding. Different product category.

How do I add Phototology to my AI agent as a memory layer?

Three integration paths: MCP Server (for Claude Code, Claude Desktop, Cursor, VS Code Copilot), TypeScript SDK (for Node agents), REST API (any language). Typical agent flow: query the registry first via GET /v1/lookup?hash=sha256(image), use the cached result if present (instant, free), call POST /v1/analyze with desired lenses on cache miss (stored automatically for future queries). See /developers for full examples.

Why not just use Claude's built-in vision for my agent?

Claude Vision is powerful, but every call costs tokens and latency, results may change if the model updates, and there's no shared memory across runs or agents. Phototology is the persistence layer: analyze once, look up forever (free), deterministic results, shared registry so other agents benefit. Best practice: use both. Phototology for persistence and decision-making, Claude Vision for exploratory/dynamic analysis.

Can multiple agents or teams share the same registry?

Yes. The registry is designed to be shared. When you analyze a photo, the result becomes available to other queries (including other agents and teams) for free. You can opt out per-analysis with privacy controls (Sovereign, coming H2 2026). This shared approach is the 'Wikipedia of images' — collective visual intelligence that compounds as more people contribute.

What happens when a new lens is added to an existing registry entry?

Existing lens results stay frozen. New lenses are appended without re-running or modifying the prior output. Your registry entry grows richer over time. This is the Accretive property: only adds, never subtracts.

Can I read the analysis without calling the API?

Yes. Phototology embeds the full analysis into the image file's EXIF/IPTC/XMP metadata. Any EXIF-compatible tool (Lightroom, Preview, ExifTool, etc.) can read the findings. The photo carries its own intelligence — and the registry lookup key — so you can rebuild the full structured record without API access.

Where it's going.

  • Collective
  • Clustered
  • Temporal
  • Portable
  • Sovereign
  • Composable

Start building.

Start building

1,000 free analyses per month. Lookups always free. No credit card.

Phototology | Analyze. Structure. Remember. | Phototology