kalibr 1.0.26

Multi-Model MCP SDK — deploy to GPT, Claude, Gemini, Copilot from one codebase.

Kalibr SDK v1.0.26

Multi-Model AI Integration Framework

Write once. Deploy anywhere. Connect to any AI model.

Kalibr turns your Python functions into APIs that automatically work with GPT Actions, Claude MCP, Gemini Extensions, and Copilot Plugins — all from one codebase.
It’s the unified SDK layer for building, packaging, and deploying AI-ready endpoints.

---

🧠 Core Purpose

Kalibr is a multi-model integration SDK that converts simple Python functions into fully MCP-compatible APIs for every major AI model ecosystem.

One function → Four schemas → Deploy anywhere.

---

⚙️ What It Does

1. Unified Schema Generation

Kalibr automatically generates and serves schemas for:

• /openapi.json → GPT Actions
• /mcp.json → Claude MCP
• /schemas/gemini → Gemini Extensions
• /schemas/copilot → Copilot Plugins
• /models/supported → List of supported integrations

No manual YAML or JSON schema creation needed.

---

2. Environment-Aware Base URLs

Kalibr auto-detects where it's running and sets the correct base URL automatically:

Environment	Example Base URL
Local	http://localhost:8000
Fly.io	https://<app>.fly.dev
Render	https://<app>.onrender.com
Custom	Set via KALIBR_BASE_URL

---

3. Deployment & Runtime Abstraction

Kalibr provides a single CLI entrypoint for local and hosted runtime deployment:

kalibr serve my_app.py
kalibr deploy my_app.py --runtime fly|render|local
kalibr version

Each runtime automatically generates valid schema URLs and deployment bundles (Dockerfile, fly.toml, etc).

---

4. Two Development Modes

Simple Mode (Function-Level)

For lightweight APIs and test integrations.

from kalibr import Kalibr

app = Kalibr(title="Weather API")

@app.action("get_weather", "Fetch weather data")
def get_weather(city: str):
    return {"city": city, "temp": 72, "condition": "sunny"}

Advanced Mode (Full App)

from kalibr import KalibrApp
from kalibr.types import FileUpload, Session

app = KalibrApp(title="Document API")

@app.file_handler("analyze_doc", [".pdf", ".docx"])
async def analyze_doc(file: FileUpload):
    return {"filename": file.filename, "result": "parsed"}

@app.session_action("save_data")
async def save_data(session: Session, data: dict):
    session.set("data", data)
    return {"saved": True}

Includes:

• Async/await support
• File uploads
• Session persistence
• Streaming responses
• Workflow scaffolding

---

5. Built-in Routes

Endpoint	Purpose
/health	Health + version check
/docs	Swagger UI
/models/supported	Shows model compatibility
/openapi.json, /mcp.json, etc.	Model schemas

---

6. CLI Reference

kalibr serve my_app.py        # Run locally
kalibr deploy my_app.py       # Deploy to Fly/Render
kalibr examples               # Copy examples
kalibr version                # Show SDK version

---

7. Value Proposition

For developers or MCP infrastructure projects (e.g. Emergent):

• Instant MCP onboarding — one file → all model schemas
• Zero config — no schema or deployment setup required
• Multi-runtime support — local, Fly, Render, or custom hosts
• Unified interface layer — consistent schema output for all AI platforms

---

8. Upcoming Additions

• Observability + tracing hooks
• Usage metering and billing
• Schema diffing / auto-validation
• Multi-runtime load routing

---

License

MIT License © 2025 Kalibr Team
GitHub: https://github.com/devonakelley/kalibr-sdk

---

Kalibr SDK — the unified layer between AI models and the real world.
Write once. Deploy anywhere. Integrate everything.