Developers

Developer portal

Everything documented about the layer, with an honest status on each part. Data model and data categories are real and complete. The public interface is in development, and is labelled as such wherever it appears.

SectionStatusWhat it covers
Getting startedpartialWhat exists, what does not, and how to evaluate the data today.
Data modelliveThe canonical entities: Asset, MarketSnapshot, SocialSnapshot, SentimentSnapshot, NewsItem, Narrative, EcosystemSnapshot.
AssetsliveIdentity, provider mapping, categories, channel pointers.
Market dataliveFields, cadence, coverage and limitations.
Social dataliveActivity series and channel series, and the difference between them.
SentimentliveDefinition, windows, aggregation levels, limits.
NewsliveAggregation, deduplication, entity mapping, the rolling window.
NarrativesliveHourly topic snapshots, dynamic categories, momentum.
EcosystemliveDaily project series and their curated coverage.
AuthenticationplannedKey issuance and scopes. Not available; no self-serve sign-up exists.
Rate limitsplannedQuotas and headers. Will be published when there is an endpoint to limit.
ErrorsplannedStable machine-readable error codes.
WebhooksplannedPush delivery for data events. Not implemented.
SDKsplannedClient libraries. Not implemented.
MCP serverplannedAgent tool surface. Not implemented.
API referenceplannedEndpoint documentation. Follows the API itself.
ChangelogplannedStarts with the first public release.

Getting started

There is no key to obtain yet, so "getting started" means something different from usual: it means working out whether this data fits your problem before either side invests in an integration.

  1. Read the data model. The canonical schema is the contract. If the entities do not fit your domain, nothing else will.
  2. Check cadence against your requirement. The freshness table is measured, not aspirational. If you need sub-minute data, this layer is the wrong tool and we would rather you find that out on page two.
  3. Check coverage against your universe. Broad hourly coverage, curated daily coverage. The numbers are on each data page.
  4. Look at the application.moonboard.ai is the layer rendered. It is the fastest way to judge data quality without an integration.
  5. Get in touch with the specific shape you need: contact@neob.ai.

Integration principles

These hold regardless of how you eventually consume the data, and following them now will save a migration later.

  • Key on the canonical asset id. Not the symbol, not the name — here is why.
  • Store the timestamp with the value. A number without its observedAt cannot be validated later.
  • Model nullability. Nullable fields are documented per category and they are null in practice, not theoretically.
  • Keep cadences separate. Do not upsample a daily series into an hourly table.
  • Handle gaps. Absence of a snapshot is a normal condition.

Webhooks

Not implemented

There is no webhook delivery today — no event types, no signing scheme, no subscription management. Push delivery only makes sense once there is a public interface to authenticate against, so it follows the Data API.

The event shapes it would plausibly carry — an observation written for a watched asset, a narrative crossing a source-count threshold, a sentiment series crossing a bound, a data category falling behind its cadence — are derivable from what the layer already computes. None of them exist as a deliverable event yet, and none are documented as if they did.

SDKs

Not implemented

No client libraries are published. A generated client is only worth shipping against a stable public contract, which is what the API work has to establish first.

MCP server

Not implemented

There is no Moonlytics MCP server. Exposing the layer as agent tools is a direction we take seriously — the reasoning is set out under AI agents and designing APIs for AI agents — but nothing is installable today, and it will not be listed as available until it is.

Changelog

A public changelog begins with the first public interface release. Until then, the meaningful record of change is this documentation, and every measured figure on the site carries the date it was measured (2026-08-31 for the current set).

Free developer tools

Two tools are available now and run entirely in the browser against published data — no key, no account: