01
Authority outside the model
Models propose. Maple identities, capabilities, policy, approvals, and deterministic services decide what may happen.
Maple AI OS
Maple AI OS
The accountable operating layer for AI agents that act — model-independent authority and evidence.
Run GPT, Claude, Gemini, Hermon, or local models. Maple keeps identity, memory, permissions, deterministic execution, and signed receipts under durable operating authority — local-first and provider-independent.
Any model, local-first
Deterministic authority
Open protocols
Signed, verifiable receipts
Maple Brain Lab · DOGMA
Genomic organization as a new computing architecture
Evo Trainer develops DOGMA, the DNA-Organized Genomic Model Architecture: a compact research model built around genomic embeddings, regulation gates, structured memory, and evidence-gated evolution.
Checkpoint101
Parameters29.83M
Quality gateEvaluating
The Maple Niche
Frontier models supply intelligence. Maple supplies durable control.
Frontier labs keep shipping better models. Maple focuses on the layer they do not own for you: portable, accountable operation across providers and over time, with explicit authority, signed receipts, and evidence-gated change.
02
Continuity across model changes
Change a provider without losing agent identity, user-owned memory, schedules, policy history, receipts, or rollback state.
03
Open ecosystem, governed edges
Connect MCP tools, A2A agents, OpenAI-compatible models, enterprise APIs, and local services behind the same capability boundary.
The Operating Loop
Intelligence moves fast because authority stays deterministic
Maple turns probabilistic model output into typed, authorized, provable, reversible system change.
1
Propose
An engine (local Hermon, llama.cpp, Ollama, or cloud) drafts a typed proposal; a deterministic planner is the safe fallback.
2
Authorize
Guard validates and the fail-closed Kernel authorizes by identity, capability, policy, and approvals.
3
Execute
A deterministic executor performs only the authorized consequence — the LLM never executes.
4
Prove
WorldLine seals signed lifecycle evidence. Independent reproduction remains a separate protocol preview.
5
Evolve
Evals gate promotion, canaries limit exposure, and failed candidates remain out of the serving path.
MapleAI Ecosystem
One OS, a focused AI ecosystem
Maple AI OS is the common authority layer. Hermon supplies local inference, Wish supplies coding agents, and Maple DNA and OpenIBank prove the OS in demanding domains.
Authority and continuityMaple AI OS
The accountable AI operating system for identities, capabilities, memory, policy, deterministic execution, receipts, and evolution.
Inference engineHermon
An independent local inference engine and multi-protocol gateway for open-weight GGUF models.
Research trainingDOGMAQuality-gated DOGMA training, evaluation, checkpoints, and research serving from Maple Brain Lab.
Coding agentWishThe coding-agent experience, governed by the Maple AI OS.
DNA computing stackHermon DNAA simulation-first DNA computing stack, guarded by the Maple AI OS.
Financial OS distributionOpenIBankFinancial workflows powered by Maple-governed models and Maple commitment controls.
Open Interoperability
Keep the best of every AI platform without surrendering control
Maple is not another frontier model or agent framework. It is the stable operating boundary around them.
GPT · Claude · Gemini · Hermon · LocalModels
Route by quality, privacy, latency, cost, and policy.
Maple model policyMCP · native drivers · enterprise APIsTools
Expose only explicit capabilities with scoped credentials.
Maple capability gateA2A · Wish · custom agentsAgents
Give every agent identity, schedule, memory, and health.
Maple agent controlCloud · edge · local · regulated systemsExecution
Separate cognition from deterministic consequence.
WorldLine receiptsAcademy
Domain knowledge for builders, operators, and researchers
The MapleAI Academy turns OS design, DNA computing, finance operations, and coding-agent practice into lessons, labs, output contracts, and model-evaluation targets.
AI-native OS
How Maple keeps authority outside every model.
Open courseHermon DNA
DNA encoding, BERT, DNABERT, simulation, and safety.
Open courseHermon Finance
OpenIBank risk, reconciliation, audit, and approval gates.
Open courseHermon Code
Wish coding agents, repository patches, tests, and rollback.
Open courseAI OS Demos
See Maple AI OS drive real domains
Each domain has a narrow job. Maple AI OS supplies shared authority, receipts, and evolution; the Hermon engine supplies local inference.
Interactive previewOS Operations
Plans OS-level agent actions as proposals that Maple can authorize, execute, and record.
Maple AI OSModel previewCode AgentPowers repository changes, code review, CI repair, and deployment plans for Wish.
WishResearch demoDNA ComputingExplains DNA-computing workflows, experiment plans, and safety gates under Maple governance.
Maple DNAApplication previewAgent FinanceSupports OpenIBank analysis, policy explanation, and auditable financial workflows.
OpenIBankUse the best AI. Keep your operating authority.
Start with one governed agent, one capability boundary, and one receipt trail. Expand without tying your operating system to a single model provider.