You are evaluating the architecture
Self-hosted developer preview
Build the source, run the quickstart, and inspect the resulting receipts.
Products
Maple AI OS is the MapleAI product. Hermon, Wish, and OpenIBank have separate product boundaries. DOGMA and Hermon DNA are open research. Each line below states what can be used now and what remains unproven.
Portfolio
Maple AI OS is MapleAI's flagship. Hermon and Wish are independent Hermon products. OpenIBank is a governed finance application. DOGMA and Hermon DNA are research programs with public evidence gates.
Put deterministic authority, policy, receipts, and evaluation around model-proposed actions.
Flagship governed agent runtime
Serve local GGUF models through OpenAI, Anthropic, and Ollama-compatible protocols from one Rust binary.
Local inference engine and AI gateway
Bring terminal, editor, review, and coding-agent workflows into one native development workspace.
Local-first agentic development environment
Apply governed agents to analysis, mandates, reconciliation, and evidence-heavy financial operations.
Governed financial-agent application
Test whether genomic organization, regulation, and recursive evaluation can support useful learned computation.
Non-transformer genomic architecture research
Evaluate transformer-based biological sequence encoders against explicit genomic tasks and real-data anchors.
Genome-native encoder research
A public page is not the same as a production service. Open the readiness record for working scope, current limits, and evidence.
Review product readinessWhere To Start
Maple does not require an ecosystem-wide commitment. Start at the smallest layer that resolves a real constraint.
Inside Maple AI OS
Registry, Guard, Kernel, and WorldLine stay inside the operating system vocabulary. They are not separately purchasable cloud services, and their maturity is governed by the Maple AI OS status.
Interprets intent and selects intelligence without receiving permission to execute consequences.
Validate capabilities, policy, budgets, approvals, and commitments through a fail-closed authority boundary.
Preserve signed proposals, decisions, executions, outcomes, and rollback history as verifiable receipts.
Describe agent, policy, model, and driver artifacts through explicit schemas and versioned package contracts.
Prove the Maple lifecycle around one consequential workflow. Add adjacent products only where they create measurable value.