Metadata-Version: 2.4
Name: laissez
Version: 0.0.1a3
Summary: Agents spending you can trust
Author-email: Matthew Collins <matthew@laissez.xyz>
Requires-Python: >=3.10
Requires-Dist: eth-account>=0.13.7
Requires-Dist: fastmcp>=2.12.4
Requires-Dist: httpx>=0.28.1
Requires-Dist: x402>=0.2.1
Provides-Extra: examples
Requires-Dist: dedalus-labs==0.1.0a9; extra == 'examples'
Requires-Dist: fastmcp>=2.12.4; extra == 'examples'
Requires-Dist: openai>=2.3.0; extra == 'examples'
Requires-Dist: pydantic-ai>=1.1.0; extra == 'examples'
Description-Content-Type: text/markdown

# Laissez

Agent spending you can trust.

## What you get
- **Paid MCP server**: Wrap FastMCP tools with `create_paid_mcp_app()` so every call flows through Laissez billing.
- **Paying MCP client**: Use `LaissezClient` as the HTTP transport for MCP agents so they authenticate and settle usage automatically.
- **Transparent settlement**: Payments are finalized with x402 and appear in the Laissez App dashboard at <https://laissez.app>.

## Set up once
- `uv sync --group examples` to pull the example dependencies.
- Grab a Laissez API key via the CLI wizard when you first run the examples, or from <https://laissez.app>.
- Export your OpenAI API key (the client agent calls an OpenAI model).
- Optional: add the keys to a `.env`; the examples load it with `python-dotenv`.

## Run the paid MCP server
- Command: `uv run examples/mcp/paid-mcp-server.py`
- What happens: spins up a FastMCP app on `http://127.0.0.1:8000`, generates a temporary wallet, and exposes paid tools (e.g. `roll_die`, `multiply`) behind Laissez enforcement.

## Run the paying MCP client
- Ensure the server is live or update `BASE_URL` in `examples/mcp/mcp-client.py` to point at your deployment.
- Command: `uv run examples/mcp/mcp-client.py`
- What happens: creates a `pydantic_ai` agent backed by `LaissezClient`, pays for two dice rolls plus a multiply call, and prints the result.

## Keep exploring
- See `examples/mcp/README.md` for a deeper walkthrough of the server and client.
- Tweak prices, add more tools, or swap in different MCP agents to prototype your own billing flows.