Python 3.10+ · Zero dependencies .syn

Syntecnia

The language built for AI agents

A complete programming language where security, observability, multi-agent coordination, and LLM integration are first-class primitives. Zero dependencies. Python 3.10+.

Get Started View on GitHub

Open Source MIT License Active Development

1234567891011121314intent: "Fetch and analyze customer data"

require net("api.example.com")
require db("./store.db")

task analyze(data)
    let trends be analyze data for "patterns"
    each trend in trends
        log "Found: " + trend
    give trends

serve on 8080
    route "GET /insights"
        let data be fetch("https://api.example.com/data")
        give analyze(data)

Why Syntecnia

Everything your agents need,
nothing they don't

Built-in primitives that eliminate boilerplate and reduce attack surface — from day one.

Security by Default

Zero access until declared. Every I/O operation requires explicit capability grants via require — network, filesystem, database, and shell are all opt-in.

Multi-Agent Coordination

Spawn agents with isolated capabilities. Share state via blackboard. Coordinate workflows with signals, channels, and structured message passing.

LLM as a Primitive

analyze, decide, generate — LLM operations are native keywords, not library calls. Switch providers without touching your logic.

HTTP Server Built-in

serve on PORT with routes, auth, rate limiting, SSE streaming, templates, and content negotiation — no framework configuration required.

Human in the Loop

approve, confirm, ask, show — built-in primitives for human oversight of agent actions. Keep humans in control.

Observable by Design

trace, log, measure, checkpoint — every operation is traceable with rich error diagnostics and structured output.

Quick Install

Up and running
in seconds

Syntecnia runs on any system with Python 3.10 or later. No virtual environments, no dependency hell — just install and start building.

✓ Python 3.10+ ✓ Zero dependencies ✓ Cross-platform ✓ .syn files

pip install syntecnia

syntecnia run hello.syn

Or run from source: github.com/kitecosmic/Syntecnia ↗

Examples

See it in action

From a two-line hello world to a full agent pipeline.

Hello, World

The simplest Syntecnia program. No imports, no boilerplate — just a log statement and you're running.

Indentation-sensitive syntax
log is a built-in, no import needed
String concatenation with +
Run with syntecnia run hello.syn

# The simplest Syntecnia program
log "Hello, world!"

# Variables and expressions
let name be "Syntecnia"
let version be 1
log "Running " + name + " v" + version

AI Agent with Human Oversight

An agent that reads a file, uses an LLM to analyze it, and asks for human approval before taking action — all in ~15 lines.

require grants capabilities explicitly
analyze is a native LLM keyword
approve pauses for human confirmation
when / otherwise for branching

intent: "Review and act on report data"

require file("./reports/")
require llm("gpt-4o")

let report be read("./reports/latest.txt")
let summary be analyze report for "key risks"

show summary
let ok be approve "Proceed with mitigation plan?"

when ok
    log "Executing mitigation steps..."
    # ... further actions
otherwise
    log "Action cancelled by operator."

HTTP API Server

Spin up a full HTTP server with typed routes and JSON responses. No framework, no middleware config — serve on PORT handles it.

serve on PORT starts the HTTP server
route defines method + path handlers
fetch makes HTTP requests natively
Responses are returned with give

intent: "Serve an AI-powered insights API"

require net("api.openai.com")
require net("api.example.com")

task get_insights(query)
    let raw be fetch("https://api.example.com/data?q=" + query)
    let result be analyze raw for "trends and anomalies"
    give result

serve on 8080
    route "GET /insights"
        give get_insights("default")

    route "GET /insights/:q"
        give get_insights(params.q)

Browse all examples