Skip to main content
Specado
Spec-First Prompts · Git-native

From Fragile Scripts to Bulletproof Specs

The professional toolkit for reliable LLM engineering. Define, test, and run provider-neutral workflows with a high-performance Rust core, and use them from Python, Node.js, or the CLI with full transparency.

spec.yaml
version: "1"
metadata:
  name: lead-qualifier
messages:
- role: system
  content: |
    Qualify this lead based on their inquiry.
    Respond with QUALIFIED or NOT_QUALIFIED.
- role: user
  content: "I'm interested in your enterprise solution"
sampling:
  temperature: 0.7
  seed: 42
response:
  format: text
strict_mode: Warn

Tired of Fragile, Untestable AI Code?

Scattered Logic

Prompt strings, API keys, and model parameters are spread all over your codebase.

"Guess-and-Check" Testing

Reliably testing LLM outputs is a nightmare, making CI/CD impossible.

Framework Lock-in

Building with one agentic framework means a total rewrite if you want to switch.

The Solution

“Specado centralizes your entire LLM workflow—prompts, models, and tests—into a single, elegant spec.yaml or spec.json file. It's version-controllable, human-readable, and machine-executable.”

How It Works

A simple, powerful workflow for reliable AI engineering.

1. Define

Craft your entire prompt workflow—messages, model settings, and behavior—in a single, human-readable YAML file.

spec.yaml
version: "1"
messages:
- role: system
  content: "Qualify this lead..."
- role: user
  content: "I'm interested in your enterprise plan."
sampling:
  temperature: 1.5 # Value is intentionally out of range
strict_mode: Warn

2. Validate

Instantly check your spec for syntax errors and schema violations. Catch mistakes before they become runtime problems.

terminal
$ specado validate --spec ./spec.yaml
✅ Prompt spec is valid

3. Preview

See the exact, translated payload that will be sent to the provider API, plus a detailed report on any data loss or changes.

terminal
$ specado preview --prompt spec.yaml --provider provider.yaml
=== Translated Request ===
{
  "model": "gpt-5",
  "messages": [...],
  "temperature": 1.0
}
=== Lossiness Report ===
[WARN] Clamp: Value 1.5 above maximum 1.0 (sampling.temperature)

4. Execute

Run your spec from the CLI or integrate it into your application using the native Python and Node.js libraries.

integration.py
import specado
from specado.types import Prompt, Message


# Define and run a prompt programmatically
prompt = Prompt(
    messages=[
        Message(role="system", content="You are a helpful assistant."),
        Message(role="user", content="What is the capital of France?")
    ]
)


# Run against a provider and model by name
result = specado.run(
    prompt,
    provider="openai",
    model="gpt-5"
)


print(result.content)
Full Transparency

Never Fly Blind: Understand Every API Call

Provider APIs are not created equal. When you define a prompt with a 0.7 temperature and the provider only supports 1.0, what happens? Specado shows you. The lossiness report gives you a clear diff of what was changed, dropped, or clamped during translation.

Lossiness Report

[WARN] Clamp: Value 1.5 above maximum 1.0 (sampling.temperature)

[WARN] Drop: Parameter not supported by provider (sampling.top_k)

[INFO] Relocate: System message relocated to provider-specific location (messages[0])

Debug with confidence
The Ejection Seat

Start Simple. Scale Anywhere. Never Get Locked In.

Specado is the perfect starting point. Its provider-neutral format gives you full portability. When your project's needs change, you can inspect the exact provider payload and migrate your logic without vendor lock-in.

Specado
Specado
LG
CA
LangGraph / CrewAI
terminal
$ specado preview --prompt spec.yaml --provider provider.yaml
No vendor lock-in, ever

Core Benefits

Built for developers who demand reliability, performance, and flexibility.

✅Validated & Reliable
Validate specs against provider capabilities and schemas to catch errors before you run. Ensure your configurations are solid and prevent runtime surprises.
🚀Rock-Solid Rust Core
A high-performance Rust core means speed and memory safety you can rely on, with lightweight native bindings for your favorite language.
🐍Write Once, Run Anywhere
Write a spec once and run it anywhere—from the CLI for quick iteration, or via native Rust-powered bindings in your Python or Node.js applications.
✨Intelligent DevEx
Get rich, location-aware error messages and a detailed lossiness report on every run. Stop guessing what the provider is *actually* receiving.
🕊️No Lock-In, Ever
A provider-neutral spec means you can swap backends from OpenAI to Anthropic or others without rewriting your core logic. You own the prompt, not the provider.

Ready to Build Better AI?

python
$pip install specado
node.js
$npm install specado
rust
$cargo install specado