The harness router, one layer of the NetShow platform
The harness router: any harness, one alive face.
NetShow is the platform for agentic AI, and the harness router is the layer that lets its alive agents use any harness, ours included. Talk to one NetShow avatar. It leads your AI team, routes each task to the right model, tool or harness, asks before it spends or acts, and shows you receipts.
How the router picks a model Approvals and receipts Bring your harness home
The problem
Too many agent tools, and no one in charge.
Every new agent tool brings its own model, its own settings and its own bill. Nobody decides which one should do the job, nobody checks before money is spent, and nobody can show you afterwards what happened.
The leader you talk to
One face leads your AI team.
You talk to one avatar, with a face and a voice, and it hands each job to the right member of your AI team. You do not learn a new tool for every task: you say what you need, and the leader tells you who is doing it and what happens next.
-
Voice first
Mr. NetShow is an AI agent who greets you first. You can interrupt him anytime. Your microphone opens only with your permission.
-
Choose your leader's voice in build
Setting the leader's voice, name and manner for your business is being built now.
The router
The right model for each task, under a daily cost ceiling.
The harness router picks a model per task from the frontier families, with open-weight models built in, so a quick answer does not pay for a heavy model and a hard problem gets a strong one. Every choice stays inside a daily cost ceiling.
- Claude
- GPT
- Gemini
- Llama
- Mistral
- Qwen
Ceilings
Every priced lane runs under a daily spending ceiling. When a lane reaches its ceiling it stops for the day instead of spending more, and a closed lane stays closed.
Approvals and receipts
It asks before it spends or acts, and it keeps the receipts.
-
The approval queue
Anything that spends money or goes out in your name waits in one queue until a person approves it.
-
Run history
Every run leaves a receipt: what was asked, which model and tool did it, and what it cost.
-
Company memory
Your AI team works from one shared memory of your business, so the next task starts from what the last one learned.
Our team sets these up for you on the NetShow platform as part of your build.
Bring your harness home
Keep the agent tools you already use.
OpenClaw, Hermes, Claude Code and the CLI coders, n8n, LangGraph and others can work under the same leader, router, approvals and ceilings. Connecting one yourself is in build unless its card below says it is ready.
Our team runs these for you today.
-
BetterClaw
in build -
CellCog
in build -
Claude Code
Ready to connect -
Headless CLI agent
Ready to connect -
Codex
Ready to connect -
DeepSeek Harness
Ready to connect -
Gemini CLI
Ready to connect -
Google ADK
Ready to connect -
Grok
in build -
Hermes
Ready to connect -
Instinct
in build -
LangGraph
Ready to connect -
Muse
in build -
n8n
Ready to connect -
OpenAI Agents SDK
Ready to connect -
OpenClaw
Ready to connect -
OpenAI hosted agent
Runs at OpenAIConnected. I can use a browser at OpenAI when you say OK.
Proof
We run NetShow on it.
NetShow is built by our own AI agent team working on the platform through the harness router, under the same approvals and ceilings described on this page.
4995 reviewed units shipped by our AI agent team in the 7 days to 2026-09-30 How we count
For developers
Read the machine surfaces.
- Developer guide
- Documentation
- MCP manifest: the tools NetShow describes to other agents.
- A2A agent card: how other agents find and reach NetShow.
- API reference: every public endpoint, the harness relay included.
- Harness guide: how to bring the agent you already use, step by step.
Developer guides
Which packages include the harness router
The harness router comes with these AI employee packages on the NetShow platform.
-
AI Specialist
Market or code, with your approval
-
AI Team
Run a small team of AI employees
-
Enterprise AI Workforce
Your own AI workforce on your own server
-
Custom
Something bigger or different
By quote, after the build call.