corailtech/labcraft

A graphical dataflow programming language and IDE, built in pure Rust.
Designed from the start to be written, run and fixed by people and AI agents alike. Ambition:
to do at least what LabVIEW does, open and cross-platform.
Specification ·
Visual notation ·
Examples ·
Try it ·
Contributing
Status (2026-11-04): the first milestone in Rust has started, and nothing is usable in Rust yet.
The language and its text format are described (spec v0), with example programs. What runs today is
a set of throwaway Python tools, including a proof of concept that checks and runs the examples
against simulated instruments, and that an AI agent drives over MCP. The first milestone (M0: reading,
checking and writing programs, then running them, with a read-only diagram viewer alongside) is
described in docs/roadmap.md.
## One program, two views
The error wire (red) orders the actions and carries the error: if the measurement fails, the
nodes after it are skipped, except visa.close, marked always, which still closes the session
(blue). Values carry their physical unit ([V]).
The diagram is drawn from a text file, which is the format itself,
examples/readvoltage.lcf:
text
labcraft 1
use core
use scpi
use visa
/// Reads the DC voltage of a multimeter.
fn readvoltage(resource: string, err: error = none) -> (voltage: f64[V], err: error) {
dmm, e1 = visa.open(resource, timeout: 2 s, err: err)
raw, dmm2, e2 = scpi.queryf64(dmm, "MEAS:VOLT:DC?", err: e1)
#close: e3 = visa.close(dmm2, err: e2)
voltage = core.withunit(raw, [V])
err = e3
The diagram was rendered from this file by a throwaway tool
(tools/lcf-figcraft), following the draft
visual notation; the editor does not exist yet.
## An AI agent at work
An agent (Claude Code) writes a new function from a one-line request, using the proof of concept
over MCP: check finds a real error, the agent fixes it, the diagram is drawn from the text, and
the function runs against a simulated power supply.
## Try it
The proof of concept is in Python (3.11 or later):
sh
pip install -r tools/lcf-poc/requirements.txt
python tools/lcf-poc/lcf.py check examples
python tools/lcf-poc/lcf.py run examples checkvoltage --sim instruments/dmm-230V.toml --arg resource=SIM::DMM0
check reports errors with their line and node; run executes a function against a simulated
instrument and prints its outputs (here Verdict.Pass, 230.06 V). To let an AI agent drive it over
MCP, see tools/lcf-poc/README.md.
## Documents
The language
| Document | Content |
| docs/language-spec-v0.md | Language and file format, v0 (first draft): types and units, dataflow, errors, structures, grammar, open questions |
| docs/visual-notation.md | How a diagram is drawn: nodes, terminals, wire families, structures (first draft) |
| docs/cli.md | The command line: fmt and check, their reports and diagnostic codes (first draft) |
| examples/ | Example programs (.lcf functions, .lct types), from a multimeter check to a test station state machine |