Wire the stack you already use
Connect LLMs, shell commands, Python or Node scripts, HTTP APIs, files, Git, containers, and complete subflows.
Visual AI workflow builder
Connect AI models, scripts, files, APIs, and agent CLIs on one canvas. Debug every step, then run the same flow from the desktop, terminal, or MCP.
Early MVP · Windows desktop · local-first
Built for real work
Knodr gives developer automation a shape you can inspect. See the whole system, step through a single node, and keep every flow versioned beside your code.
Connect LLMs, shell commands, Python or Node scripts, HTTP APIs, files, Git, containers, and complete subflows.
Independent branches execute concurrently. Route, merge, loop, delay, and trigger work without serializing the whole graph.
Pause on breakpoints, inspect inputs and outputs, replay one node, and follow live logs without adding print statements everywhere.
Launch flows manually, on timers, from webhooks, file changes, or Git events—and keep long-running sessions alive.
The built-in MCP server exposes catalog, authoring, validation, execution, logs, snapshots, and node-level debugging tools.
No hosted control plane, accounts, or telemetry. Credentials remain encrypted locally and never serialize into a flow file.
One core, three surfaces
The desktop app, headless CLI, and MCP server all use the same scheduler, serializer, validator, and node registry. A flow behaves the same everywhere.
GUI
Compose on a canvas, run live, inspect values, and replay nodes.
CLI
Validate and run versioned flows in terminals and CI pipelines.
MCP
Let coding agents author, organize, execute, and debug flows through tools.
$ cat src/auth.cs | knodr run review.knodrflow \ --output-format json ✓ input.manual ready ✓ ai.claude completed 2.4s ✓ agent.codex completed 2.8s ✓ logic.merge completed 0.0s ✓ file.write review.md flow completed in 2.8s
Get Knodr