Build AI systems visually. Run them in your browser.
Most frameworks hide agent logic inside black-box nodes. NodeFox puts the whole thing on a visual canvas — connect models, files, APIs, code, and decisions, then run them as deterministic workflows, batch automations, or simple apps.
Most frameworks bury agent behavior inside individual nodes. NodeFox externalizes the entire logic onto the canvas — so the graph itself is the agent you can see, inspect, and change. It compiles to JSON and runs on a local Rust/WASM engine with real parallelism, native loops, and activation edges.
01
Global tooling
Tools — JavaScript functions or MCP servers — live at the workspace level, not locked inside one node. Any Conversation node can reach them.
02
Visual control flow
Multi-agent handoffs and conditional routing are built on the canvas with Decision nodes. Inspect and change the logic visually — no code spelunking.
03
Deterministic execution
The same inputs always produce the same routes and node sequence, so every run is repeatable, debuggable, and fully replayable.
The Node Lexicon
Learn 10 Nodes. Build Anything.
Most visual graph tools force you to memorize a sprawling library of specialized nodes. NodeFox uses just 10 composable core nodes — enough to model any pattern your team needs.
Build the system once. Publish it as a form-based App so teammates can upload files, enter inputs, click Run, and receive generated outputs in a live chat interface.
Contract analyzer
Upload a contract and get back risk flags, plain-language summaries, and the clauses that matter.
Meeting brief generator
Drop in a messy CSV and receive a normalized, schema-validated export ready to use.
Blog post generator
Enter a topic and outline to generate full posts with citations and source links.
Research report builder
Provide a question and get a multi-source synthesis with quotes, citations, and references.
Support ticket classifier
Paste a ticket and route it to the right team with a confidence score and reasoning.
Invoice extraction app
Upload invoice PDFs and extract line items, totals, and vendor metadata into structured data.
Sensitive FileReaderEnterprise OpenAIConversationSave to ExcelWriter
Browser-native by design
NodeFox agents run on your machine, inside your browser. Your workspace, files, prompts, API keys, and AI responses stay under your control. Connect your own provider keys, work offline after loading, export everything, and optionally grant access to local files or shell commands when needed.
✓Private by default — No NodeFox server processes your data — execution stays on your device.
✓BYOK providers — Connect OpenAI, Gemini, Claude, Grok, Mistral, or custom providers with your own keys.
✓Offline-capable — Build, edit, and manage networks after the app loads — no constant server round-trips.
✓Local files & shell — Optional advanced access for power users who need to read local files or run shell commands.
✓Exportable workspace — JSON and ZIP export plus local folder sync — your workspace is yours to move.
Governable in production
✓Approval gates — Hold high-impact actions for human sign-off.
✓Run history — Every step recorded — branch, data, and cost.
✓Export trail — Hand auditors the full record on demand.
Traditional frameworks vs NodeFox
NodeFox flips the traditional AI framework on its head. Instead of hiding agent behavior inside individual nodes, the entire logic lives on the visual graph — so the network itself is the agent.
Traditional frameworks
Agent behaviorHidden inside one node
Tool accessLocked to specific nodes
Routing & handoffsCode workarounds
ExecutionCloud, server-side
DebuggingComplex and opaque
NodeFox
Agent behaviorThe whole graph is the agent
Tool accessAgent creation with scoped tooling
Routing & handoffsVisual Decision nodes
ExecutionLocal-first, your own keys (BYOK)
DebuggingDeterministic & replayable
Plans for builders and production teams
Choose the plan that fits your team size and workflow volume.
Answers about who it's for, how it works, and how it compares.
Who is NodeFox for?
NodeFox is built for teams that need AI agents to be understandable and governable after launch. That includes product engineering, applied AI, platform, security, operations, and business systems teams that share responsibility for reliability.
Is NodeFox in beta?
Yes — NodeFox is in beta and evolving fast. We recommend a staged rollout: start with bounded agents, keep high-impact actions behind approval gates, and expand as your team validates results.
How does NodeFox differ from Zapier?
Zapier is excellent for straightforward trigger-action automations. NodeFox focuses on multi-step graph orchestration with richer branching, stateful patterns, and AI-centric control flows.
How does NodeFox differ from n8n?
n8n is a flexible agent tool with visual automation and code extensions. NodeFox is purpose-built for AI orchestration — with predictable execution, native MCP tool access, and step-level run inspection and approval flows.
Can NodeFox run without full autonomy?
Yes. Most teams run a tiered model: low-risk paths auto-complete, medium-risk paths route to review, and high-risk paths require explicit approval. This keeps throughput high while reducing the chance that ambiguous model outputs trigger irreversible actions.
Where do MCP capabilities fit?
MCP gives agents standardized access to external tools and context while keeping you in control of every route. Teams can scope MCP usage by branch, attach validation checks to tool outputs, and maintain a full audit trail of what was called and why.
Build agents your team can trust
Start with visual orchestration, scale with predictable control, and ship AI systems your whole team can understand in production.