Kalibr

Your agent workflows shouldn't be a black box.

Deploy MCP servers in one line. Trace every call. See the full causal chain.

Know exactly what your agents cost.

Open-source SDK for instant MCP deployment. Enterprise observability to track costs and trace workflows across GPT, Claude, Gemini, and every AI model.

Get Started (Open Source) Contact Sales (Observability)

Get Started

Install Kalibr and deploy your MCP server to all AI models instantly.

1. Install

pip install kalibr

2. Create your first app

Create a new file called my_app.py and add:

from kalibr import Kalibr

app = Kalibr(title="My API")

@app.action("get_data", "Fetch user data")
def get_data(user_id: str):
    return {"user_id": user_id, "status": "active"}

Or run the included demo:

kalibr examples
kalibr serve kalibr_examples/basic_kalibr_example.py

3. Run it

kalibr serve my_app.py

Your API is now live with multi-model schemas for all platforms:

http://localhost:8000/openapi.json   # GPT Actions
http://localhost:8000/mcp.json        # Claude MCP
http://localhost:8000/schemas/gemini  # Gemini Extensions
http://localhost:8000/schemas/copilot # Microsoft Copilot
http://localhost:8000/docs            # Interactive docs

That's it. One Python file — fully MCP-ready for GPT, Claude, Gemini, and Copilot.

Beyond Deployment: Full Observability

The open-source SDK deploys your MCP server. Enterprise adds complete visibility.

Open Source (Free)

✓ One-line MCP deployment
✓ Multi-model compatibility
✓ Automatic schema generation
✓ Local development tools

pip install kalibr

Enterprise Observability

⚡ Complete workflow tracing
💰 Cost attribution per call
🔍 Cross-vendor visibility
📊 Real-time analytics dashboard

Contact Sales →

See the Complete Picture

[Your App] →$0.12→ [GPT-4] →$0.08→ [Claude] →$47.00→ [Your MCP Tool]
                                        ↑
                              Cost explosion detected

When your AI workflow costs explode, Kalibr shows you exactly where, why, and how much each step costs. Available with Enterprise Observability.

What can Kalibr do for you?

Start with open-source deployment. Scale with enterprise observability.

OPEN SOURCE

Deploy MCP servers instantly

One SDK connects to GPT, Claude, Gemini, and Copilot. Write once, deploy anywhere.

OPEN SOURCE

Automatic schema generation

Kalibr generates GPT Actions, MCP schemas, and platform-specific definitions automatically.

ENTERPRISE

Track costs across every vendor

See exactly what each model, tool, and API call costs. Know where your AI budget goes.

ENTERPRISE

Debug cross-vendor workflows

When GPT calls Claude calls your tool, see the complete execution path and find failures instantly.

The Problem We Solve

73% of AI companies use multiple models.
0% can trace what happens between them.

Every workflow now crosses vendors: GPT → Claude → Tools → APIs.
But logs stop at boundaries. Costs can't be attributed. Failures can't be debugged.

Kalibr is the only observability layer built for multi-model AI.

Start free with open-source deployment. Add observability when you need to see what's really happening.

Curious about Kalibr?

How does Kalibr relate to MCP?

Kalibr deploys MCP servers instantly and adds observability on top. MCP defines how tools are exposed; Kalibr ensures they work everywhere and shows you what's happening.

What's free vs paid?

The core SDK for MCP deployment is open source and free forever. Enterprise observability (cost tracking, cross-vendor tracing, analytics) requires a license.

Which platforms are supported?

GPT, Claude, Gemini, and Copilot are fully supported. Mistral and others are in development.