UNIDROIT Principles 2016
RemoteAbout UNIDROIT Principles 2016
UNIDROIT Principles 2016 packages fixed legal texts as executable rules for users asking closed-form legal questions. Answers include grounds and source text, and are checked against scenarios.
Key features
- 01Pinned UNIDROIT source text
- 02Topic-specific rule packages
- 03Scenario-based answer checks
- 04Hosted or local evaluation
Use cases
- 01Assessing contract formation questions
- 02Analyzing limitation and non-performance
- 03Examining hardship and interest cases
MCP support
It offers a hosted MCP server and a local Docker Compose setup for MCP clients to submit questions. Both use the same pinned canon and engine.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.arxo.io/mcp/piccclaude mcp add --transport http io-arxo-picc https://mcp.arxo.io/mcp/picc{"mcpServers":{"io-arxo-picc":{"type":"http","url":"https://mcp.arxo.io/mcp/picc"}}}Open in Cursor{"servers":{"io-arxo-picc":{"type":"http","url":"https://mcp.arxo.io/mcp/picc"}}}Open in VS Code{"mcpServers":{"io-arxo-picc":{"type":"http","url":"https://mcp.arxo.io/mcp/picc"}}}{"mcpServers":{"io-arxo-picc":{"serverUrl":"https://mcp.arxo.io/mcp/picc"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
20| Parameter | Type | Required | Description |
|---|---|---|---|
| kinds | array | No | narrow the kinds of candidate |
| limit | integer | No | how many candidates (8, at most 25) |
| query | string | Yes | a question or a notion in your own words |
| package | string | No | search inside this instrument only |
Details
From the project READMEarxo-unidroit-picc
Pinned canon, executable as it stands: source text with provenance, the
model built over it and the scenarios that check its answers. The norms
here are data — they are executed by law; this repository carries no
rule code of its own.
Contents
| Package | Version | Scenarios | Dependencies |
|---|---|---|---|
unidroit.picc | 0.1.0 | 5 | 0 |
unidroit.picc_formation | 0.1.0 | 4 | 1 |
unidroit.picc_limitation | 0.1.0 | 3 | 1 |
unidroit.picc_non_performance | 0.1.0 | 4 | 1 |
unidroit.picc_obligations | 0.1.0 | 4 | 1 |
unidroit.picc_performance | 0.1.0 | 4 | 1 |
unidroit.picc_validity | 0.1.0 | 4 | 1 |
How it is verified
Every package is built by package.py with the pinned toolchain: compilation,
§267 scenarios on both evaluators, replay of saved evaluations and a rebuild
from the release itself. A failure of any step stops the build.
python3 tools/toolchain.py download --lock toolchain.lock.json --out /tmp/law-tools
python3 /tmp/law-tools/package.py packages/<name> \
--lawc /tmp/law-tools/law-cli --offline --out /tmp/<name>-release
Dependencies
Dependency bytes are pinned under packages/<name>/deps/ and verified against the contentHash recorded in law.lock: the build and the CI are offline.
Asking these packages
Two ways, the same slice of the canon and the same engine; the difference is whose machine evaluates.
- Hosted:
https://mcp.arxo.io/mcp/picc— requests are evaluated on Arxo infrastructure. - Local:
docker compose -f mcp/compose.yml up— nothing leaves the machine.
mcp/pin.json records what either one must answer from: the
content hash of the canon profile, the pinned packages, the ABI
version of the engine and the availability of the hosted endpoint.
An answer that cannot be traced to that pin is not this repository's
answer.
Worked cases
Each collection is a case package of its own: the facts, the questions and
the answers, with the pinned canon as its dependency. Ask one from its
directory with law ask . — nothing is downloaded.
Rights
Code and model: see LICENSE. The pinned texts of the publishers keep their
own terms of reproduction, listed per package in NOTICE; those terms are
not covered by the code licence.
Maintainer: Rifat Dzhumagulov ([email protected]). Address: https://github.com/arxohq/arxo-unidroit-picc.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.