Every enterprise arrangement is different. These are the building blocks — most clients combine the API layer with a custom archive build.
Every colour entry in the archive was hand-researched and written from a named primary source. This is what makes Colour Memory different from every other colour API.
Primary sources only
Every archive entry cites a real, named, verifiable source — museum records, literary letters, dye chemistry journals, court documents, natural history texts. No Wikipedia. No aggregated colour databases. If a source can't be named, the entry is flagged as insufficient evidence.
Hand-written at entry level
Every archive entry was individually researched and written — not generated, not scraped, not licensed. The colour notes average 220 words per entry and include the historical consequence of the colour, not just its appearance. This depth is what makes semantic search return genuinely useful results.
Evidence-graded outputs
Every result carries a claim_role (anchor, contextual, or reject) and a claim_strength grade. Anchor entries have direct institutional or documentary evidence. Reject entries are stubs — the API flags them and tells your system not to use them in client-facing work.
Colour science, not just naming
The archive feeds a deterministic colour science layer: CIEDE2000 perceptual distance, Lab/LCh/LRV metrics, WCAG 2.1 contrast ratios, palette role assignment, and cultural risk scoring by market. These computations run without any LLM. Results are identical on every call.
No hallucination by design
Colour Memory does not generate colour information from training data. It retrieves it from the archive, computes it from colour science, or returns an explicit insufficient-evidence signal. The do_not_say field on every result tells your agent exactly what claims the evidence does not support.
Bespoke archives use the same standard
Custom enterprise archives are built to the same evidence standard as the core archive. Every entry we write for your institution or domain has a named primary source, a consequence note, and a claim role. You get verifiable, citable colour intelligence — not a content dump.
01 · Most popular
White-label API access
Full Colour Memory API served under your domain or embedded in your product. Your users never see the Colour Memory name unless you choose to credit it.
Custom subdomain or direct integration
Your branding on all responses and PDFs
Volume pricing from 50,000 calls/month
SLA with guaranteed uptime
Dedicated technical support
02
Bespoke archive build
A custom colour archive built specifically for your domain — your collection, your period, your material culture. Evidence-grade entries, named primary sources, the same schema as the main archive.
Custom archive scoped to your subject
Evidence-grade entries with primary sources
Integrated into the API alongside or instead of standard archives
Ongoing expansion on retainer if needed
03
Embedded intelligence layer
Colour Memory as a background service in your existing workflow. Your design system, your CMS, your AI agent — all gaining access to named, sourced colour matching without surface-level integration.
MCP server or REST integration
Works with Claude, ChatGPT, and custom agents
No user-facing UI required
Results carry provenance and do_not_say guardrails
04
Research & editorial partnership
For institutions, publishers, and broadcasters who need colour intelligence for specific projects. Archive research, palette builds, cultural risk assessment, and editorial colour reporting.
Project-scoped archive research
Palette PDF deliverables with source citations
Cultural risk assessment for specific markets
Credited or uncredited depending on brief
These are representative examples of the kinds of products enterprise access supports.
Heritage institution
Collection colour search for a national museum
Researchers query the museum's holdings by colour, period, and material. The archive returns named, sourced colour entries matched to the collection's own digitised objects.
Design agency
Brand audit tool sold to luxury clients
An agency embeds the brand.audit endpoint in their own client portal. Clients upload palettes, receive cultural risk assessments and archive provenance, billed as a proprietary agency service.
SaaS platform
Colour naming for an e-commerce platform
A product photography platform adds archive-derived colour names to every image processed. The ecommerce.namer endpoint runs in the background — users see intelligent colour names, not hex codes.
Publisher
Editorial colour intelligence for a style title
A magazine uses the archive as a research tool for colour features. Journalists query historical colour meanings, cultural risk, and provenance — results are verified and citable.
AI product
Colour layer for a creative AI assistant
An AI creative tool integrates Colour Memory via MCP. When users ask about colour choices, the agent has access to historically grounded, culturally aware colour intelligence rather than generating responses from training data alone.
Interior design platform
Heritage palette engine for a materials marketplace
A materials platform surfaces historically grounded colour stories alongside products. Customers see not just a colour name but its archive entry — the period, the source, the consequence.
Standard plans give full API access. Enterprise removes limits and adds customisation.
The API is standard REST with JSON responses. MCP integration takes minutes. Custom deployment options on request.
Authentication
API key via X-Api-Key header. Enterprise keys are rate-unlimited and can be scoped per environment.
Endpoints
65+ REST endpoints. All also available as MCP tools via a single SSE URL. Dot-notation naming: archive.search, brand.audit, query.hex.
Response format
JSON with consistent structure across all endpoints. Every colour result includes hex, name, archive, primary_source, claim_role, and do_not_say fields.
Deterministic core
Colour matching, WCAG analysis, cultural risk scoring, and palette roles are fully deterministic — no LLM, no variance between calls.
Uptime
99.9% SLA available on enterprise plans. Hosted on Railway infrastructure, US West. Custom deployment to other regions on request.
Data & IP
Query data is logged for archive improvement by default. Enterprise plans can disable logging and include custom IP agreements for proprietary archive builds.