mimOE · The Agentix Operating Engine

Your AI agents.
Your hardware.

Models, agents, tools, and the mesh between your machines, running on devices you control. No cloud required, no per-token bill.

Up and running in under 5 minutes.

mimOE Studio drop a screenshot at assets/hero.png(Overview page)
Local inference
84 tok/s on your GPU
Runtime healthy
0 cloud calls$0 / token
Agents

A serious agent harness without the claw. On a local model.

Files, shell, APIs, skills, subagents: an agent that carries long multi-step work end to end, with you approving what executes.

How agents work
An agent running tools mid-task drop a screenshot at assets/agents.png(agent chat with tool calls)
Tool approval
bash · npm test
AllowDeny
Models

Pull any model from Hugging Face. Serve it by name.

Catalog → quantization → pull → loaded on your GPU. Aliases let apps ask for qwen while the router picks the model.

How models & aliases work
Models: active list and local store drop a screenshot at assets/models.png(Models page)
Serving as
qwen → Qwen3.6-35B
Providers

Ollama, vLLM, OVMS, cloud. One router over all of it.

Attach the inference servers you already run and register keyed cloud providers. Their models land in one list, behind one router, with real measured tok/s.

How providers work
Providers: attached Ollama server and a cloud provider drop a screenshot at assets/providers.png(Providers page)
ollama · live
qwen3.6:35b63.6 tok/s measured
Playground

Try any model. Instantly.

One Try button from any model to a raw conversation. Thinking streams in, tokens-per-second on every turn, nothing saved. Need the ceiling instead of a single turn? The built-in benchmark swarms the box to measure how many agents it can host.

About the Playground
The Playground: try a model drop a screenshot at assets/playground.png(Playground mid-answer)
Throughput
84 tok/s this turn
Benchmark

How many agents can this box host?

A swarm of real agents with real sessions, driven against your instance until it bends. Density, completion rate and the breakpoint, measured by running it on your hardware, not read off a spec sheet.

About Benchmark
Benchmark: a swarm run measuring agent capacity drop a screenshot at assets/bench.png(Benchmark mid-run)
Capacity
300 agents one box
The mesh

Models on machines around you. Yours to use.

mimOE nodes discover each other. The workstation's 35B model shows up on your laptop, usable by name.

How the mesh works
Nearby models discovered on the mesh drop a screenshot at assets/mesh.png(Models → Nearby)
Routable by name
{node-id}/Qwen3.6-35B
MCP Gateway

MCP built in. One gateway for every tool.

Agents don't wire their own MCP servers. Attach any MCP server once, scope it with client keys, and every tool call flows through one governed MCP gateway.

How the MCP gateway works
MCP Gateway: an attached server with its tools and agent access drop a screenshot at assets/gateway.png(MCP Gateway page)
Every call, brokered
keys scoped per agent
Remote instances

Drive any mimOE on your network. By node ID.

Switch to another instance in one click. If its IP changes, Studio finds the node on the mesh and follows it.

How remote connections work
Manage Connections drop a screenshot at assets/remote.png(Manage Connections overlay)
Reachable · node DEMO
key accepted9 service keys found
Observability & Governance

Every inference leaves a receipt.

Each call through the router becomes an OpenTelemetry trace, capturing model, routing, tokens, timing, and content on-device. Governance, the policy and audit layer, is being built on exactly these records.

How trace & governance work
Trace: a conversation's inferences with tokens and timing drop a screenshot at assets/trace.png(Trace view)
Trace · shipped
OpenTelemetryGovernance · coming soon

Own the stack.

Models, agents, tools, mesh, and the receipts to prove what happened.

One-click runtime install includes the engine, model store, router, and a tool gateway. Works offline.