The protocol for reproducible AI workflows across all languages, systems, and models.
AI IMPROVISES └── Different outputs each run └── Results drift across versions └── Can't reproduce in production └── No audit trail └── Enterprise adoption blocked YOUR AI PIPELINE TODAY: Input → ? → ? → ? → Output?
Input → Parse → Validate → Runtime → Hash → Output
↓ ↓ ↓ ↓ ↓ ↓
SHA1 AST Rules Canon SHA256 SAME
EVERY TIME. EVERY LANGUAGE. EVERY SYSTEM.
# Install in 60 seconds
curl -sSL https://scl-lang.org/install | sh
# Write your first SCL
cat > hello.scl << 'EOF'
SCL:V1
handles {
name("input")
greeting("output")
}
scl {
"Hello, " @name "! Welcome to deterministic AI."
}
EOF
# Run it
scl run hello.scl
# Verify determinism
scl hash hello.scl
# Output: 7d4e3eec80026719639ed4dba68916eb94c7a49a053e05dcb5732a2a602ce0c6
# This hash NEVER changes
SCL:V1
handles {
input("data")
validated("clean")
output("result")
}
scl {
"Validate " @input " → "
@validated " → Process → "
@output
}
SCL:V1
handles {
template("doc_type")
data("json")
document("output")
}
scl {
"Generate " @template
" from " @data " → "
@document
}
SCL:V1
handles {
task("request")
agent1("analyzer")
agent2("executor")
result("response")
}
scl {
@task " → " @agent1
" → " @agent2 " → "
@result
}
scl parse
scl validate
scl run
scl hash
Python SDK
Go SDK
JavaScript SDK
TypeScript SDK
VSCode Extension
Dev Server
Linter/Formatter
Visualizer
Takes 60 seconds. Zero dependencies. Works everywhere.
Install SCL Now →Enterprise? Contact for deterministic AI assessment →