Running in minutes, not theory.

# 1. Get the code (Rust required: https://rustup.rs)
git clone https://github.com/adubinsky/devlish
cd devlish

# 2. Build and add `devlish` to your PATH
./install.sh

# 3. Write a program
cat > hello.dvl << 'EOF'
Ask "What is your name?" as user name
Print user name
EOF

# 4. Run it
devlish run hello.dvl --input '{"user_name": "World"}'

# Also useful:
devlish validate hello.dvl                    # check syntax
devlish compile hello.dvl -o hello.dvlc.json  # compile to bytecode
devlish disassemble hello.dvlc.json           # inspect the bytecode

The run command auto-compiles .dvl files in memory; devlish script.dvl is shorthand for devlish run script.dvl.

# 1. Start the MCP server over a folder of .dvl tools
devlish mcp --tools-dir ./tools/

# 2. Register it with your MCP client, e.g. Claude Code:
claude mcp add devlish -- devlish mcp --tools-dir /path/to/tools

# 3. Every .dvl file is now a callable tool. This IS the tool:
Ask "Customer tier?" as customer_tier
Ask "Deal size?" as deal_size

discount_rate equals 0.05
If customer_tier is "enterprise":
  discount_rate equals 0.15

Respond with record with deal_size times discount_rate as discount

Ask lines define the tool's input schema. Respond with returns typed JSON; Fail with returns a structured error the LLM can parse and retry.

# 1. Install the runtime
npm install devlish-runtime

# 2. Compile a rule with the CLI (or commit the .dvlc.json)
devlish compile pricing.dvl -o public/rules/pricing.dvlc.json

// 3. Load and run it in your app
import { loadTool } from "devlish-runtime";

const tool = await loadTool({
  bytecode: await fetch("/rules/pricing.dvlc.json").then(r => r.json())
});

const result = await tool.run({ customer_tier: "enterprise", deal_size: 500000 });
console.log(result.response); // => 75000

Programs cannot reach your network or filesystem; tools that declare HTTP or file permissions are rejected at load time.