Agentic IDE · Production-Grade · Enterprise-Ready

Build agents workflows guardrails evaluations notebooks that ship to production.

An agentic IDE built from scratch. Prototype to production-ready services with guardrails, multi-model evals, and one-click export — in minutes, not months.

Get Started View on GitHub

Terminal — syntaX

# Clone and start all 7 services $ git clone https://github.com/dshakes/syntax && cd syntax $ make run-local Starting syntaX data plane... ✓ gateway :8080 ready ✓ model-router :8081 ready ✓ guardrails :8082 ready ✓ simsearch :8083 ready ✓ eval-harness :8084 ready ✓ export :8085 ready ✓ audit :8086 ready # Launch desktop app $ cd desktop && npm run tauri dev syntaX running at https://tauri.localhost

syntaX — Agent Editor

Entrypoint

Router Agent

Guardrails

Writer Agent

MCP Server

Drag nodes

Drag & drop to build agent graphs

Node Config

Model claude-sonnet-4.5

Temperature 0.7

Policy Enterprise

Status Active

Why syntaX

From prototype to production in minutes, not months

⚖️

Benchmark before you ship.

Compare accuracy, latency, cost, and safety across models side-by-side. Built-in evaluation harness with DLP scanning and guardrails enforced on every request. Not opt-in — mandatory.

🔧

Visual-first, code-ready.

Drag-and-drop agent builder that generates production Go REST APIs. No vendor lock-in. Route to any model from any provider with a single click.

📦

Your infrastructure. Your rules.

Deploy on your infrastructure. Export to Docker, Helm, or Fargate. Every component independently deployable behind your own trust boundaries.

See it in action

Purpose-built for production AI

Every screen ships with production features. See multi-provider model management, guardrail testing, evaluation benchmarks, and one-click export — all in real time.

syntaX — Models

Model Endpoints

3 Active

OpenAI

gpt-4.1

v: latest128k ctx

active

✓ 200 OK — 38ms

Anthropic

claude-sonnet-4.5

v: latest200k ctx

active

✓ 200 OK — 52ms

Huggingface

llama-3.3-70b

v: latest32k ctx

active

✓ 200 OK — 67ms

Guardrails

6 Active Rules

Enterprise Internal

Blocks PII, scans for secrets, enforces content guidelines

Maximum 5/5 rules

Open Research

Permissive policy for research workloads

Permissive 1/4 rules

Live Testing Console

✗ PII detected — email address found

⚠ Sensitive content — personal data reference

✓ No prompt injection detected

✓ No secrets or API keys found

Evaluation Results

Run Complete

Accuracy

94.2%

Avg Latency

340ms

Safety Score

98.1%

gpt-4.194.2%340ms$0.4298.1%

claude-sonnet-4.592.8%280ms$0.1899.2%

llama-3.3-70b89.5%520ms$0.0696.4%

Export to Production

Go + Docker

✓ Select

✓ Format

✓ Configure

4 Preview

5 Download

// Auto-generated by syntaX — production-ready Go service

package main

import (

"net/http"

"github.com/syntax/guardrails"

"github.com/syntax/router"

)

func main() {

r := router.New("customer-support-agent")

r.Use(guardrails.Enterprise())

r.Use(guardrails.AuditTrail())

http.ListenAndServe(":8080", r)

}

Multi-provider model registry with one-click connection testing

Capabilities

Everything you need to ship AI to production

A complete platform with guardrails built in from day one — not bolted on as an afterthought.

Visual Agent Editor

Drag-and-drop agent graphs. Connect LLMs, tools, vector stores, and custom logic as composable nodes.

MCP Playground

Connect, test, and discover Model Context Protocol tools. Integrate MCP servers into agent workflows as first-class nodes.

Model Management

Multi-provider model registry with one-click routing. Swap between OpenAI, Anthropic, and self-hosted models instantly.

Guardrails

PII detection, prompt injection defense, DLP scanning, content filtering, and custom policy rules. Enforced on every single request.

Notebooks

AI-native interactive notebooks with a built-in conversational assistant. Multi-turn streaming, code generation, and one-click apply.

Evaluations

Benchmark prompts across models. Compare accuracy, latency, cost, and safety scores side-by-side with structured test suites.

Export

One-click export to production Go repos with auto-generated REST APIs, DLP scanning, Docker, Helm charts, and Fargate task definitions.

Demo Mode

One-click demo with simulated data and pre-configured agents. Explore every feature without any API keys or setup required.

Dark & Light Theme

Beautiful theming that respects your system preferences. Carefully crafted for long coding sessions with full accessibility support.

Architecture

Three planes. Strict trust boundaries.

Every component is independently deployable. The UI never calls LLM providers directly — all inference routes through the data plane with guardrails enforced.

UI Plane

💻 Tauri v2 Desktop

⚙️ React 18 + TypeScript

🎨 Tailwind CSS

Data Plane — Go Microservices

🔗 API Gateway :8080

🤖 Model Router :8081

🛡 Guardrails :8082

🔍 SimSearch :8083

📊 Eval Harness :8084

📦 Export :8085

📜 Audit :8086

Control Plane — LLM Providers

🌟 OpenAI

🧰 Anthropic

💪 Self-Hosted

🔌 MCP Servers

Get Started

Ready to build?

Clone the repo, start the services, and build your first agent graph in under five minutes.

Download Latest Release View Documentation

$ git clone https://github.com/dshakes/syntax && cd syntax && make run-local Copy

$ cd desktop && npm install && npm run tauri dev Copy

Build agents workflows guardrails evaluations notebooks that ship to production.

From prototype to production in minutes, not months

Benchmark before you ship.

Visual-first, code-ready.

Your infrastructure. Your rules.

Purpose-built for production AI

Everything you need to ship AI to production

Visual Agent Editor

MCP Playground

Model Management

Guardrails

Notebooks

Evaluations

Export

Demo Mode

Dark & Light Theme

Three planes. Strict trust boundaries.

Built with modern, proven tools

Ready to build?