Downloads & Tools
Get Nexus Runner for local skill execution, or integrate via SDK. Pro+ subscribers get early access to Harness Agents.
Nexus Runner
Desktop app for running AI skills on your own machine.
Nexus Runner for Windows
Windows · Desktop App·v1.0.12·~139.8 MB
Graphical local execution environment. Manage and launch any installed skill directly from your desktop. Python runtime included.
System Requirements
- · Windows 10 / 11 x64
- · 4 GB RAM (8 GB recommended)
- · ~500 MB free disk space
What's Included
- ✓ Python runtime bundled — nothing extra to install
- ✓ One-click install, GUI skill manager
- ✓ Local + cloud hybrid execution out of the box
NexusRunner-1.0.12-win-Setup.exe
Windows will show a prompt on first install
The installer is not code-signed yet, so Windows SmartScreen may warn about an unknown publisher. Choose More info, then Run anyway to continue. You can verify the download against the published hash and version.
To verify: compare the installer's SHA-512 with the version info below
Nexus SDK
mcp-nexus-sdk·v1.3.2
Headless Python SDK that drives the Nexus skill library, agentic chat loop, and skills-as-tools adapter from your own client — no Electron Runner required.
Requirements
- · Python ≥ 3.10
- · Windows 10 / 11 x64
- · Nexus API key (free / Pro)
What's in 1.3.2
- ✓ Agentic chat loop with streaming and ref-table
- ✓ Skills-as-tools adapter (OpenAI + Anthropic) for your own agent
- ✓ Full skill-library control: catalog / install / run / balance / sync
Common questions
Windows says "Windows protected your PC" on first run — what now?
That is SmartScreen's routine notice for a newly signed certificate, not a malware warning. Click More info, then Run anyway. The installer's SHA-256 is listed under the download button on this page if you want to verify it yourself.
Does Nexus Runner update itself?
Yes. It checks on launch, downloads in the background, and applies the update on the next start so a running task is never interrupted. You can turn auto-update off in Settings and come back here to update manually.
Does it work offline?
Skills that execute locally keep working with no network, and your files never leave the machine. Only skills that fetch live data (search, FX rates) and account quota sync need a connection.
Inside the Runner
After installing the Runner you'll find these four core panels — together they cover everything from one-off runs to long batch jobs.
Chat · Conversational execution
Invoke any installed skill like a chat message. Describe your task in plain language and the Runner matches it to the right skill and runs it locally.
Skills · Install & manage
Browse your installed and purchased skills, one-click install / update / remove. Each skill ships with docs, parameter templates, and example inputs.
Workstation · Batch jobs
Drag files in, run pipelines across many inputs, chain skills together. Built for one-off heavy jobs like cleaning 10k rows or generating 50 decks.
Usage · Metrics & logs
Real-time view of this month's quota, execution history and error logs. Fully synchronised with your web account — consistent across devices.
Skills Ecosystem — Build, share, and reach users
MCP Nexus is not a single app — it's a community-driven marketplace of reusable skills. 38+ curated skills already cover data, content, finance, ops, and security, all vetted and one-click installable into your Runner.
• Developers: package your scripts, prompts, or workflows as .ocskill and publish to reach the whole platform • End users: can't find the skill you need? Post a request and community developers will respond • Enterprise contributors: signed skills get featured placement and priority visibility
Runner Downloaded — What's Next?
- 1 Run the installer, then launch Nexus Runner from your Start menu
- 2 Go to API Keys and create a key to connect Runner to your account
- 3 Head to My Skills, install a skill, and click Execute
Code Examples
import requests
API_KEY = "sk_live_your_api_key_here"
BASE_URL = "http://localhost:5000"
headers = {
"X-API-Key": API_KEY,
"Content-Type": "application/json"
}
def use_skill(slug: str, version: str = "1.0.0"):
resp = requests.post(
f"{BASE_URL}/api/v1/license/use",
json={"skill_slug": slug, "skill_version": version},
headers=headers
)
resp.raise_for_status()
return resp.json()
def prefetch_tokens(slug: str, count: int = 10):
resp = requests.post(
f"{BASE_URL}/api/v1/license/prefetch",
json={"skill_slug": slug, "skill_version": "1.0.0", "count": count},
headers=headers
)
resp.raise_for_status()
return resp.json()["tokens"]
def get_balance(slug: str):
resp = requests.get(
f"{BASE_URL}/api/v1/license/balance?skill_slug={slug}",
headers=headers
)
return resp.json()
if __name__ == "__main__":
result = use_skill("image-ocr")
print(f"Token expires: {result['expires_at']}")
print(f"Remaining: {result['remaining_nex']}")Harness Agents
Coming SoonAutonomous AI agents that plan, execute, and iterate across multi-step workflows — no prompts, no babysitting.
Subscribe Pro+ for Early Access
Harness Agents are available to Pro and Enterprise subscribers first. Upgrade today to join the beta waitlist.
ChatBI Agent
Data Analysis · BI
Local AI-powered BI tool. Connect your data sources, auto-generate charts and reports, and query in plain language — no SQL required. Data never leaves your machine.
McKinsey Agent
Strategy · Consulting
AI-powered strategic consulting desktop tool. Structured business analysis with McKinsey-style insight reports and actionable recommendations.
OpenStock Agent
Finance · Investment
Local stock analysis engine with real-time Finnhub market data. AI-driven deep analysis and investment insights stored in local MongoDB — fully private.
RPA Agent
Finance · Process Automation
RPA automation purpose-built for finance workflows. Covers OCR, bank reconciliation, voucher generation, audit workpapers, budget analysis, and 10+ core skills with scheduled task support.
Workflow Agent
Process Design · Automation
Standardized flowchart generation and management. Supports Visio-standard nodes, Mermaid conversion, multi-format export, and AI-powered diagram generation and audit.