Open app
Moonborn — Developers

Moonborn v1.0 — general availability

The four-layer persona engine, the six-step pipeline, six SDKs, an OpenAI-compatible endpoint, an MCP server, and 153 endpoints across 17 feature families. v1.0 is the contract.

After eighteen months of building, Moonborn v1.0 is generally available. This post is the contract: what's stable, what's still landing, what you can build on today.

What's stable

Everything under /v1/ of the API. 153 operations across 17 feature families:

  • Personas — generate, refine, fork, audit, test, fingerprint, versions.
  • Chat — sessions, messages, streaming, drift detection, long-term memory.
  • Marketplace — publish, install, fork, review, collections.
  • Webhooks — 16 event types, HMAC signed, retried, replayable.
  • Billing — subscriptions, invoices, usage, plans.
  • Config — 8-layer resolver, snapshots, rollback, audit-trail.
  • Privacy — DSAR exports, account deletion, retention policies.
  • Auth + Identity + SCIM — credentials, MFA, SAML, SCIM 2.0.
  • API keys — scoped tokens, IP allowlist, rotation.
  • Admin — platform staff endpoints.
  • OpenAI-compat — drop-in /v1/openai/chat/completions.
  • MCP — Model Context Protocol server.
  • System — health, version, OpenAPI.

The path-segment version is the contract. v1 minor + patch bumps are additive only (see API versioning policy).

What's preview

Two surfaces ship behind a "preview" label — the application logic is live, the public HTTP surface is one release behind:

  • Git sync (persona-as-code). Domain and application layers are done; the /v1/integrations/git/* REST endpoints + onboarding UI land in v1.1. Early-access customers can wire it up via direct integration today.
  • Marketplace commerce. The payout pipeline (Stripe Connect, 70/30 split) is gated behind marketplace.commerce.enabled, off by default. Enterprise customers can enable it; broader rollout in v1.2.

We're explicit about preview so you don't bet roadmaps on shapes that will move. Things called "preview" will move.

What you can build today

The set of products we already see customers shipping:

  • Character-consistent support agents — drift detection + long-term memory + brand voice persona.
  • Brand voice variants at scale — one canonical persona, locale and tone forks, lineage-tracked.
  • Multi-character scenes — ensemble relationships, multi-persona sessions, distinctiveness floors.
  • Synthetic user panels — 5-10 persona panels, structured interviews, longitudinal consistency.
  • OpenAI-stack migration — base-URL swap, model = persona ID, streaming + tools pass through.
  • MCP-driven IDE integration — personas as MCP resources, chat as a tool.

What's next

The roadmap is short:

  • v1.1 (next month) — Git sync HTTP API + onboarding UI. More per-audience tutorials. TR + DE translation passes.
  • v1.2 (Q3) — Marketplace commerce broader rollout. Custom moderation classifiers per workspace. More language SDKs.
  • v2 (eventually) — when we have enough additive load that we need a clean break. Today: not soon, and not planned. v1 has runway.

Thank you

To the early-access cohort who flagged the edge cases, broke the pipeline in productive ways, and pushed us to write the audit + drift machinery instead of shipping naive prompt-stuffing. You know who you are. The docs are better for it; the product is sharper for it.

Onward.