Clinical terminology mapping you can actually measure.
Codon is a FHIR R4 terminology platform that tells you how reliable each mapping is, keeps you in control of every AI suggestion — or runs 100% manually, with no AI at all — manages your terminology licences, and comes as a managed European-cloud service, on-prem optional. Official content built in — bring or build your own alongside it.
Measured reliability · AI optional, or fully manual · licences managed · official content, plus your own.
Try it yourself, right here.
Demo queries are throttled to 5 per 30 seconds by a browser-side courtesy limit.
For developers
Every request above is a standard FHIR R4 API call. Copy the curl command, or explore the full API in our interactive documentation.
Every mapping tells you how far to trust it.
Codon grades each mapping on the strength of its specific source→target pair and shows that grade wherever the mapping appears — so a clinical-safety or procurement reviewer can judge the risk before a code is ever used. A reliability number always comes with the method behind it.
Strong, Review-recommended, or Manual — graded for the exact source and target systems in play, and shown wherever the mapping appears.
Reliability is tested against an independent, published benchmark, versioned, and refreshed with each release. A number never travels without its method.
Reliability is a property we test — and publish.
Match strength, one pair at a time
SNOMED CT→ICD-10 is a hard, lossy crosswalk; LOINC→NPU is close to exact. Codon grades each pair on its own terms — Strong, Review-recommended, or Manual — instead of one blanket accuracy claim.
Benchmarked against public ground truth
Every reliability score is measured against an independent, published benchmark — a shared public standard. Evidence you can verify for yourself, rather than a figure to take on faith.
An evidence report, not an adjective
Reliability ships as a downloadable report per mapping set — the evidence artifact clinical-safety and procurement reviewers ask for, ready to attach to a submission.
A full FHIR terminology platform — governed, not magic.
Terminology operations
Look up, validate, translate, and expand clinical codes across SNOMED CT and every terminology you load — through a standards-based FHIR R4 API.
Cross-system translation
Translate a code between two loaded systems through their concept maps. When there's no direct map, Codon chains existing maps through intermediaries — every hop is a real crossmap, never an inferred one — and if no such chain exists, it returns no match rather than inventing one.
AI-assisted mapping, human-in-the-loop
The system proposes; your terminologist approves, rejects, or adjusts. Nothing auto-accepts, every suggestion is marked as AI-generated — and the AI is optional, so you can map entirely by hand if you prefer.
De-identification
Detect and pseudonymise patient identifiers in clinical free text before it moves — processed in-region, never leaving the platform.
Encounter validation
Check encounters against registry rules and extract codes before submission — catch problems upstream of the registry.
Versioning & audit
Versioned releases, full change history, and a tamper-evident audit trail — so every change to meaning is accountable.
One platform, a lot of very different jobs.
Teams reach for Codon across problems that look nothing alike. A sample of what we've seen so far:
Terminology as an API
A product team calls one standards-based API to look up, validate, translate, and expand codes — instead of building and running terminology infrastructure in-house.
validation · translationEmbed the terminology layer
A platform vendor puts code search, text-to-code suggestions, and cross-system translation inside their own interface, with the terminology layer running behind it.
extraction · translationValidate before submission
A data team checks encounters against national registry rules and resolves the flagged issues before submitting.
validationDe-identify clinical text
A team pseudonymises patient identifiers in clinical notes for a data-reuse workflow, with the AI running in their own environment and text never leaving it.
de-identificationCode free-text notes
Coders receive diagnosis and procedure suggestions read from clinical notes — each a reviewable draft a person confirms, with negation and timing flagged.
extractionAuthor & release official content
An authority maintains its own classification in a private workspace, cuts a versioned release, and publishes it with append-only history and a tamper-evident trail.
authoringTrack licensed-vocabulary entitlements
A safety and standards team tracks entitlements, usage, and audit across its own licensed clinical vocabularies — in one place.
licence managementSee what a version change breaks
A team compares two releases of a vocabulary concept by concept and identifies what changes in its data before the upgrade lands.
mappingTranslate decades of registry data
Researchers translate historical registry data from legacy ICD revisions through ICD-10 and ICD-11 to SNOMED CT, using published intermediary maps, and export the harmonised set.
translation · mappingMap to a common data model
An evidence team maps internal source codes to the OMOP common data model with AI-assisted suggestions a person reviews — keeping a record of every decision.
mappingModernise legacy messages
An integration project converts legacy clinical messages into profile-conformant FHIR, with terminology validated in flight rather than after the fact.
validation · translationA terminology backend
A third-party platform calls Codon in production to validate and translate codes behind its own product.
validation · translationA cross-section, not a limit — the same operations recombine into many more. Tell us the job you have in mind.
Built for the EU AI Act, not retrofitted to it.
Codon treats AI as something you have to be able to trust and answer for — with the governance controls that health-data buyers are increasingly audited against.
- ✓AI is optional. Switch it off entirely and run Codon 100% by hand — every terminology and mapping operation works without it.
- ✓Output marking. AI-generated mappings are labelled in-band in the data model and carry into the Mappings CSV export, with machine-readable export marking for the remaining formats following the harmonised Article 50 format.
- ✓No silent automation. Suggestions never auto-accept; a person signs off on every one.
- ✓Abstention. When the model isn't confident enough, it declines to suggest rather than guessing.
- ✓Accountable by design. A tamper-evident audit log records who changed which mapping, and when.
You decide where your patients' data lives.
By default, Codon runs as a managed service in the European cloud — EU-owned and in-region, outside US jurisdiction for your clinical data. Prefer another region, or to keep everything inside your own environment? That's a supported option, not the only path.
Official content, plus your own. The built-in catalogue holds authority-published terminology and mappings only — nothing algorithmically derived is passed off as canonical. Alongside it you can manage your own content — custom code systems, value sets, and maps, plus opt-in derived crosswalks — kept tenant-private and clearly separated from the official catalogue.
Global standards, with the deepest Nordic coverage.
These figures count the official, authority-published built-in catalogue only — SNOMED International, NLM, WHO, Regenstrief, and national authorities; your own custom content is additional and not counted here. Shown live from the catalogue; not padded with derived or duplicate rows.
Pricing
Start with the free tier. Upgrade when you need more.
Foundation
Try the API and build a prototype
- Look up and check codes through the FHIR API (read-only)
- Read access to the code systems your licence permits
- 1,000 API calls a month
- No credit card required
Insight
For teams using the API in production
- Everything in Foundation
- Translate codes, including multi-step, and run advanced concept queries (ECL)
- Build and manage your own value sets and mappings
- AI query assist: plain language to concept queries, with explanations
- Single sign-on (SSO)
- 1 million API calls a month; CSV and HL7v2 export
Clinical
For clinical coding, hospital IT, and research teams
- Everything in Insight
- All AI mapping methods, with a human review queue
- AI text coding that understands negation and time
- Author codes, set up workflows, manage releases
- 10 million API calls a month, all 9 export formats
Enterprise
For EHR vendors, regions, and national agencies
- Everything in Clinical
- Your own dedicated environment
- Run it on your own servers if you prefer
- Custom SLA and call volume
- Patient-data processing included
Compare plans
| Foundation | Insight | Clinical | Enterprise | |
|---|---|---|---|---|
| Price | €0 | €299/mo | €1,499/mo | Custom |
| Monthly ops budget | 1K | 1M | 10M | Custom |
| Rate limit (sustained) | 10 req/s | 50 req/s | 200 req/s | 500 req/s |
| FHIR R4 read ($lookup, $validate-code, $autocomplete) | ✓ | ✓ | ✓ | ✓ |
| $expand, $closure, $batch, $subsumes | — | ✓ | ✓ | ✓ |
| $translate + multi-hop translate | — | ✓ | ✓ | ✓ |
| ECL queries | — | ✓ | ✓ | ✓ |
| ValueSet authoring + mapping workspace | Read-only | ✓ | ✓ | ✓ |
| FHIR resource suite (forms, knowledge artifacts, profiles, care pathways) | — | — | ✓ | ✓ |
| Clinical text coding | — | — | ✓ | ✓ |
| AI mapping (suggest, infer, cross-lingual, analogical) | — | — | ✓ | ✓ |
| AI query assist (NL→ECL, explain) | — | ✓ | ✓ | ✓ |
| AI editorial assist (release notes, scope notes, retirement impact) | — | — | ✓ | ✓ |
| Code set management (authoring, workflows, releases) | — | — | ✓ | ✓ |
| Terminology wheel & corpus analytics | — | — | ✓ | ✓ |
| Terminology sharespace (share value sets, mappings, supplements) | — | — | ✓ | ✓ |
| Export formats | — | CSV + HL7v2 | All 9 | All 9 |
| Format conversion (HL7v2, CDA, IPS) | — | — | ✓ | ✓ |
| Bulk data export | — | — | ✓ | ✓ |
| FHIR Subscriptions | — | — | ✓ | ✓ |
| SSO / OIDC sign-in | — | ✓ | ✓ | ✓ |
| PHI processing (LPR validate, extract-codes, de-identify) | — | Add-on | Add-on | Included |
| On-premise deployment | — | — | — | Optional |
| Custom code systems | — | — | ✓ | ✓ |
| Soft-cap overage | — | — | — | ✓ |
| Custom SLA & uptime | — | — | — | ✓ |
| Dedicated onboarding & support | — | — | — | ✓ |
PHI processing is a contracted add-on enabled with a signed DPA, and is included with Enterprise. On-premise deployment is available on Enterprise.
Paid plans include a 14-day trial. PHI processing is available as an add-on with a signed DPA.
Contact
Have questions or need help choosing the right tier? Send us a message. We respond within one business day.
Codon is in a controlled evaluation phase. Request access
See how reliable your mappings really are.
Tell us the code systems and mappings you work with. We'll show you the measured-reliability view, the governance controls, and a sample evidence report to take away.
Available in English, Danish, and Norwegian — with more languages as we grow.