AgeniusDesk: Every automation, one desk.

One desk, every capability

Everything you need to run n8n at scale.

AgeniusDesk · Executions / Errors

DIAGNOSE

Failures, grouped and explained.

Every error from every instance in one feed, grouped by workflow. One click asks AI for the root cause and a suggested fix.

AgeniusDesk · Observe

OBSERVE

See exactly where a run slowed.

Per-execution OpenTelemetry traces from n8n. Node-by-node timing pinpoints the bottleneck or the failure, not just that something broke.

AgeniusDesk · Code Lab

BUILD

Describe it. Ship it to n8n.

Code Lab is a Monaco editor with an AI assistant. Generate Code-node logic or a whole workflow, then push it straight into your instance.

AgeniusDesk · Models

MODELS

Your models, your keys, per task.

Code Lab, error triage, and chat each get their own provider and model, with fallback. OpenAI, Anthropic, OpenRouter, Perplexity, DeepSeek, Mistral, local Ollama, and more.

AgeniusDesk · Secrets

SECURE

Secrets stay server-side.

A Fernet-encrypted vault stores your secrets and injects them as credentials into any connected n8n. Keys never touch the browser.

AgeniusDesk · Containers

DEPLOY

Stand up n8n in one click.

Spin up a fresh self-hosted n8n (plus Postgres, Redis, Qdrant, Ollama) in Docker, and manage every container from one screen.

And more in the desk

Everything else you'd wire up by hand.

Fleet Health

Workflow health and error rates rolled up across every connected instance, in one pane. A degraded instance is shown, not fatal.

Insights

Execution analytics: success rates, error trends, and your busiest workflows over time, per instance and combined.

The Harness

An Obsidian-compatible notes vault with full-text search, tags, and backlinks. Your agents’ knowledge, versioned as files you own.

n8n intelligence

n8n-mcp and a curated skill library auto-installed on first boot, so Code Lab and the assistant build workflows right the first time.

Cost tracking

LLM spend folded into the trace layer: per-trace and per-call cost, enriched from token usage against a live price book.

Accounts & roles

Local accounts with optional TOTP 2FA, CSRF protection, and role-based access, so a shared self-hosted desk stays yours.

Optional · Agent Fleet

Run your own agents, the way you run n8n.

A managed fleet of LangGraph and PydanticAI agents: a catalog, a live node graph, a normalized run waterfall, human-in-the-loop approve/resume, and LangSmith tracing with per-call token and cost. Built on the LangChain ecosystem, with PydanticAI running alongside through its own adapter. Build your own in Code Lab and register them, no restart.

Off by default — a stock install stays a pure n8n control plane. Opt in with the agent extra:

# opt in to the agent stack (LangGraph + PydanticAI) $ docker build --build-arg AGD_EXTRAS="assistant,langgraph" . # …or from source $ pip install '.[langgraph]' # force it on when the extra is present $ AGD_AGENTS_ENABLED=true

Community modules

Extend the desk. On your terms.

Install third-party modules straight from GitHub. AgeniusDesk pins the commit, scans the code, and shows exactly what it can touch before you consent, then runs it isolated if you want.

01

Inspect

Point AgeniusDesk at a GitHub repo. It pins the exact commit and runs a static scan, diffing declared capabilities (network hosts, filesystem writes, subprocess) against what the code actually does.

02

Consent

Proportional, informed consent. Critical findings make you type the module id to confirm; every install is written to an audit trail with who, when, and the commit.

03

Isolate

Run a module out-of-process: a sandboxed subprocess, or its own hardened Docker container. Privileged actions go through a loopback capability bridge, so provider keys and the host never leak.

Available now

YouTube Research v0.2.0

Paste a YouTube link, transcribe it from captions, and get a structured breakdown auto-filed into your notes vault under research/.

View on GitHub →

Proxmox v0.1.0

Connect a Proxmox VE cluster: live nodes, VMs, and LXCs with gated start/stop/reboot. The cluster token is injected host-side by the bridge, so it never reaches the module.

View on GitHub →

Build your own MIT

One repo, many modules: a manifest, a FastAPI router, and an optional sandboxed view. Ship it, and anyone installs it in two clicks.

View on GitHub →

Heuristic scan, not a sandbox by default — install in-process modules only from sources you trust. For a real OS boundary, enable the container tier.

How it compares

The only one that does all of it.

AgeniusDesk

Administrate

n8nTrace

DIY Grafana

Self-hosted

–

Open source (MIT)

–

One-click n8n deploy

–

Multi-instance fleet

–

Per-execution tracing

–

Cost observability

–

AI root-cause diagnosis

–

AI Code Lab (build + ship)

–

AI agent fleet (opt-in)

–

Community modules

–

Encrypted secrets vault

–

Integrated Docker manager

–

Price

Free

Paid

Free

Free, DIY

Quickstart

Up and running in about a minute.

$ git clone https://github.com/Mfrostbutter/ageniusdesk-ce $ cd ageniusdesk-ce $ docker compose up -d

Then open localhost:3000, create your account, and connect your first n8n.

Runs anywhere Docker Desktop Bare metal Cloud VPS

Open source

Built in the open.

Early and shipping fast, star it and help shape the roadmap.

100%

Self-hosted. Your infra, your rules.

MIT

Licensed. No fair-code limits.

0

Telemetry. Your data stays put.

★ Star on GitHub

AgeniusDesk Cloud

Want it hosted? Join the Cloud waitlist.

Same desk, fully managed. Be first when AgeniusDesk Cloud opens.