OCL Context Layer

Official registry AI and agents

Reference MCP server for OCL, an open standard giving AI agents scoped business meaning for Odoo.

Install

Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).

OCI package ghcr.io/nantiai/ocl-examples:0.1.0-alpha.1
Claude Code
claude mcp add ocl-context-layer -- docker run -i --rm ghcr.io/nantiai/ocl-examples:0.1.0-alpha.1
JSON configuration (Claude Desktop, Cursor, VS Code)
{
    "mcpServers": {
        "ocl-context-layer": {
            "command": "docker",
            "args": [
                "run",
                "-i",
                "--rm",
                "ghcr.io/nantiai/ocl-examples:0.1.0-alpha.1"
            ]
        }
    }
}

Documentation

From the project's README on GitHub, refreshed weekly. Links open on GitHub.

Contents
  1. OCL standard
  2. Use OCL
  3. Status
  4. External developer alpha
  5. Core idea
  6. Learn more
  7. Independence
  8. License

OCL standard

OCL is an experimental open format for giving AI systems structured, scoped, and verifiable business meaning for Odoo.

An Odoo connector explains how to reach records. An OCL context pack explains what relevant records and fields mean, which tempting interpretations are wrong, what ambiguity must be resolved, and which safety boundaries still apply.

question -> OCL context -> AI reasoning + existing connector -> Odoo

Use OCL

Choose the surface that fits your product:

RouteCallBenefitAlpha availability
MCP for AI agentsget_context, resolve_noun, explain_fieldPut scoped Odoo meaning directly in an agent's tool loopPublic example or invitation runtime
CLIget-context "What is revenue?"Inspect a context pack and its warnings immediatelyPublic examples now
Python validatorvalidate_entry(entry, schema)Read and produce interoperable OCL documentsAvailable now
TypeScript integrationFive production operation shapesConnect a product to the verified hosted runtimeSource alpha and invitation runtime

Clone and try the context behavior:

git clone https://github.com/Nantiai/ocl-standard.git
cd ocl-standard
python reference/python/ocl_examples.py get-context "What is revenue?"

The result asks which revenue definition applies and returns the paired warning against treating account.move.amount_total as universal revenue.

Add the example MCP to an agent
{
  "mcpServers": {
    "ocl-public-examples": {
      "command": "python",
      "args": [
        "/absolute/path/to/ocl-standard/reference/python/ocl_examples.py",
        "serve-mcp"
      ]
    }
  }
}

This local server demonstrates the real OCL tool contract using ten public candidate examples. It is not the commercial Odoo 19 runtime.

See Using OCL for the integration flow, MCP calls, CLI commands, SDK examples, response handling, and current access boundary. Connector and MCP implementers should freeze against the connector runtime contract and its conformance fixture. Hosted runtime clients can follow the verified alpha call guide. Developers can first try one live verified decision, then start a scoped seven-day trial without GitHub, package installation, Odoo credentials, or customer data. The GitHub request and email paths remain available as fallbacks.

This repository contains the open interoperability layer:

  • JSON Schemas for semantic entries and request-scoped context packs;
  • a dependency-light Python validator and TypeScript client shape;
  • ten Odoo 19 candidate examples covering the major knowledge types;
  • an executable conformance suite;
  • an example-only resolver and MCP server;
  • public benchmark methodology.

It does not contain nanti.ai's commercial registry, production assembler, verification harness, source fingerprints, private benchmarks, tenant overlays, memory, or unreleased ontology candidates.

Status

Version 0.1.0 is an experimental technical alpha. The format can change while independent implementations test it. The examples demonstrate the format; they are not the commercial verified registry and do not carry a lifetime guarantee.

Odoo 19 is the only current stock-version focus.

External developer alpha

Developers can now clone the standard, run conformance, integrate the example MCP server, and report an independent implementation. See the developer alpha guide.

The public alpha is an interoperability surface, not free bulk access to the commercial verified registry. A bounded verified runtime is live at https://api.context.nanti.ai; its native form immediately issues a scoped, expiring evaluation key.

Core idea

meaning -> evidence -> rule -> assertion -> applicability -> confidence

Facts do not become verified because an LLM wrote plausible JSON. A conforming document describes evidence and assertions. A registry operator separately decides which verification policy is sufficient for a release claim.

Learn more

Independence

Odoo is a trademark of Odoo S.A. This project is independently developed and is not affiliated with or endorsed by Odoo S.A.

License

The open repository is licensed under the Apache License 2.0. The commercial OCL registry and services are separate and are not distributed by this repository.

Frequently asked questions

What is OCL Context Layer?

Reference MCP server for OCL, an open standard giving AI agents scoped business meaning for Odoo.

How do I install OCL Context Layer?

It runs on your machine. In Claude Code: claude mcp add ocl-context-layer -- docker run -i --rm ghcr.io/nantiai/ocl-examples:0.1.0-alpha.1. For other clients, paste the JSON configuration from this page into their MCP settings.

Does OCL Context Layer need an API key?

Its registry entry lists no required key or secret. Some features may still need an account with the service it connects to.

Is OCL Context Layer open source?

Yes. Its code is on GitHub under the Apache-2.0 license.

Who maintains OCL Context Layer?

It is published by Nantiai. AI.tn lists it but does not maintain or endorse it.

From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.