Provides tools for interacting with Odoo 19's External JSON-2 API, enabling AI agents to perform operations such as search, search_read, read, create, write, unlink, and search_count on Odoo models, as well as checking server connectivity via the version endpoint.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@odoo-json2 Search res.partner for companies named Deco and show their names."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Cursor plugin that exposes Odoo 19 External JSON-2 as two MCP tools. The domain type is a single OdooCall — not one tool per model, and not XML-RPC or JSON-RPC.
OdooCall = { model, method, ids?, context?, params }
POST {ODOO_URL}/json/2/{model}/{method}
Authorization: bearer {ODOO_API_KEY}
X-Odoo-Database: {ODOO_DATABASE}Tool | What it does |
| One JSON-2 call. Named kwargs only. |
|
|
The odoo-json2 skill documents search, search_read, read, create, write, unlink, and search_count, plus the one-transaction-per-call rule.
Related MCP server: odoo-mcp
Cursor IDE (desktop). Grok Bot cannot load ~/.cursor/plugins/local. Grok Bot plugins are account-wide marketplace connectors, not this local folder.
Node.js 18 or newer (global fetch, no npm dependencies).
An Odoo 19 database on a Custom pricing plan (the external API is not available on One App Free or Standard).
A user API key and the database name.
Copy this directory to ~/.cursor/plugins/local/odoo-json2.
Reload the window (Developer: Reload Window).
Open Plugins → Configure and set the variables below.
Confirm the odoo-json2 MCP server and the odoo-json2 skill appear under Customize.
After changing plugin variables, toggle the odoo-json2 MCP server off and on. Reload Window alone does not pick up new env values.
Do not put API keys in this repo. The plugin only declares variable names.
Declared in .cursor-plugin/plugin.json and substituted into mcp.json:
Variable | Required | Meaning |
| yes | Host or origin. |
| yes | User API key (see below). |
| yes | Database name. Sent as |
mcp.json launches ./server.mjs as a plugin-relative executable. Cursor's local plugin loader does not expand ${PLUGIN_ROOT} in args.
In Odoo: Preferences → Account Security → New API Key.
Give the key a description and a duration (maximum three months). The value is shown once — copy it into the Cursor plugin variable. For integrations, Odoo recommends a dedicated bot user with the minimum access rights rather than a personal admin account.
The server sends Authorization: bearer … with a lowercase bearer, matching the Odoo 19 docs.
search_readThis request and result are the deco / company example from the official Odoo 19.0 External JSON-2 documentation (dummy host mycompany.example.com). They are not live data from this plugin.
HTTP (docs):
POST /json/2/res.partner/search_read HTTP/1.1
Host: mycompany.example.com
X-Odoo-Database: mycompany
Authorization: bearer …
Content-Type: application/json; charset=utf-8
{
"context": { "lang": "en_US" },
"domain": [
["name", "ilike", "%deco%"],
["is_company", "=", true]
],
"fields": ["name"]
}Documented success body:
[{ "id": 25, "name": "Deco Addict" }]Same call through odoo_call:
{
"model": "res.partner",
"method": "search_read",
"context": { "lang": "en_US" },
"params": {
"domain": [
["name", "ilike", "%deco%"],
["is_company", "=", true]
],
"fields": ["name"]
}
}Prefer search_read over search then read. Each JSON-2 request is its own SQL transaction.
export ODOO_URL=https://mycompany.odoo.com
export ODOO_API_KEY=your-key
export ODOO_DATABASE=mycompany
node server.mjsThe process speaks MCP over stdin/stdout (newline-delimited JSON-RPC). Logs go to stderr.
npm run check # node --check server.mjs
npm test # stdio initialize + tools/list; mocked odoo_call (no live Odoo)Tests never use a real API key or a public Odoo instance.
.cursor-plugin/plugin.json
mcp.json
server.mjs
skills/odoo-json2/SKILL.md
package.json
LICENSEMIT
Call one Odoo 19 External JSON-2 method: POST /json/2/{model}/{method}. Body is named kwargs only (ids, context, plus params). One SQL transaction per call.
| Name | Required | Description | Default |
|---|---|---|---|
| ids | No | Record ids for record methods (read, write, unlink, …). Omit for @api.model methods. | |
| debug | No | If true, include Odoo error.debug traceback on HTTP errors. Default false. | |
| model | Yes | Technical model name, e.g. res.partner | |
| method | Yes | Model method, e.g. search_read | |
| params | No | Extra named kwargs for the method: domain, fields, limit, offset, values, … Do not use positional args. | |
| context | No | Odoo context object, e.g. {"lang": "en_US"} |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden of behavioral disclosure. It discloses that it uses named kwargs and runs one SQL transaction per call, but it does not mention potential data mutation, error handling, authentication requirements, or side effects. For a generic method caller, this is a significant gap.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences with zero fluff. The main action (calling a method) is front-loaded, followed by the key request format and transaction detail. Every word earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a generic method-call tool with no output schema and no annotations, the description is too sparse. It does not explain the response format, error behavior, or how to construct complex params (e.g., domain, fields). The agent would have to infer too much from the schema alone, which is insufficient for a tool that can invoke arbitrary methods.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so baseline is 3. The description adds the note that only named kwargs (ids, context, plus params) are allowed, clarifying the relationship between parameters and going beyond the schema's individual descriptions. This raises the score to 4.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States exactly what it does: calls one Odoo 19 External JSON-2 method via a specific HTTP endpoint. Clear verb+resource, and the sibling odoo_version is obviously different, so no ambiguity.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies this is the tool for making method calls, and the only sibling is odoo_version which serves a different purpose. However, it does not explicitly state when not to use this tool or mention alternatives, so it falls short of a 5.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
GET {ODOO_URL}/web/version — connectivity check. No API key. Returns { version, version_info }.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It discloses the HTTP method, the lack of API key requirement, and the expected response structure ({ version, version_info }). This is sufficient for a simple connectivity check, though it does not mention error behavior or idempotency, which are minor omissions for a GET endpoint.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, concise sentence that front-loads the endpoint, method, and purpose. Every word adds value, with no extraneous information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a zero-parameter, read-only connectivity check, the description provides the endpoint, authentication requirement, and return fields. With no output schema, this is complete enough for an agent to invoke correctly. The low complexity means no additional context is needed.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are zero parameters, so the schema already fully covers parameter semantics (100% coverage vacuously). The baseline for 0 parameters is 4, and the description correctly omits any parameter-specific details since none exist.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states a specific verb (GET) and resource ({ODOO_URL}/web/version) with an explicit purpose of connectivity checking. It distinguishes from the sibling tool odoo_call by indicating this is a dedicated version/connectivity endpoint rather than a general call tool.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage for connectivity checks and indicates no API key is needed, which is helpful context. It does not explicitly state when not to use it or name alternatives, but the sibling odoo_call suggests a clear separation of concerns for version checks versus general calls.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.1odoo_callodoo_versionScored across 2 tools
The two tools are clearly distinct: one is a generic RPC caller for Odoo model methods, and the other is a simple version/connectivity check. There is no overlap or ambiguity between them.
Both tool names follow the consistent odoo_ prefix with lowercase snake_case. While one uses a verb (call) and the other a noun (version), the pattern is uniform and predictable.
Only two tools is thin for a server exposing an Odoo API, but the generic odoo_call tool is broad and intentional. This makes it borderline rather than severely under-scoped.
The generic odoo_call can invoke any Odoo model method, covering common CRUD and workflows, so there are no fatal dead ends. However, dedicated convenience tools for frequent operations are absent, which is a minor gap.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Odoo MCP Pack — ERP/CRM via Odoo's external JSON-RPC API.
MCP server unifying ERPs, CRMs, APIs and knowledge base for Claude, ChatGPT and Gemini.
Hosted MCP server for AI-driven data ops. Create apps, manage schemas, and CRUD structured data.
Self-hosted MCP gateway: turn any API, database or MCP server into AI connectors — no code.