MCP

The Model Context Protocol (MCP) is an open standard that lets an AI assistant read data and interact with other applications. A server exposes a set of tools, and the assistant discovers and calls them on your behalf. For Patalyze, that means pulling filings, mapping claims, and building a research for you.

Patalyze runs two MCP servers, and they do different jobs. The split is deliberate: one lets an assistant look, the other lets it act.

Start with Global Patents. It is one URL and four read-only tools, so there is nothing to configure and nothing it can break. Pulling up patents and prior art is usually the first thing you want from an assistant anyway. Add the Patalyze server once you want an agent that builds and edits inside your databases.

Either server works with any MCP client. If you use a specific assistant, the ChatGPT and Claude guides walk through setup there; the rest of this page is the reference both build on.

How connecting works#

Both servers are remote HTTP endpoints that authenticate with your Patalyze account over OAuth. There are no API keys to create, copy, or rotate.

You add a server by its URL. The client opens a browser, you sign in to Patalyze and approve the access it requests, and from then on the client holds a token scoped to your account. Each server declares the scopes it needs, listed in its section below, and an assistant can only do what those scopes allow. The servers are independent, so connect either one or both.

Global Patents MCP#

Read-only access to the global patent corpus and CPC/IPC classifications. Because it never touches your research databases, it is the safe first install: connect it and immediately ask about prior art, a competitor's filings, or what a CPC class actually covers. It stays useful alongside the Patalyze server later, for quick lookups mid-research.

It requests two scopes:

• patents:read to search and read patents.
• classifications:read to search CPC/IPC classifications.

And exposes four tools:

• search-patents searches the global corpus with text, bibliographic filters, and semantic similarity.
• get-patent returns a single patent by publication number.
• search-classifications searches CPC/IPC classifications by symbol, text, or meaning.
• get-classification returns a classification and its tree by symbol.

Under the hood this is the Data API: the four tools run the same search engine over the same global corpus that the Data API exposes over HTTP, and each maps to one of its REST endpoints. The patent filter syntax is identical, so the Endpoints reference doubles as the field and operator vocabulary for search-patents.

Patalyze MCP#

The full workbench. This server opens your research databases to an MCP client: products, patents, patent-to-product mappings, dashboards, charts, and notes. An agent acts with your permissions and can both read and edit. This is the server that lets an assistant build out a research end to end.

It requests four scopes:

• research:read and research:write to read and modify your databases.
• patents:read and classifications:read to search the global patent corpus and classifications.

It exposes 48 tools. They are grouped below by what they act on; reach for the group that matches your task.

Research databases#

• list-research lists the databases in your organization that you can read.
• start-research creates a new database.
• update-research renames a database or sets its emoji or icon.
• move-research moves a database into a folder or out to the library top level.
• delete-research permanently deletes a database and everything inside it.

Folders#

• list-folders lists the library folders and the tree of databases inside them.
• create-folder creates a folder, optionally nested inside another.
• move-folder moves a folder and its whole subtree elsewhere in the library.
• delete-folder deletes a folder and everything inside it, recursively.

Patents#

• search-patents searches the global corpus by text, bibliographic filters, and semantic similarity.
• list-patents lists the patents added to a database.
• get-patent reads one patent in a database, joined with its product mappings.
• add-patents adds patents to a database by publication number, or by a query that imports all matches.
• delete-patents removes patents from a database (the global corpus is left untouched).

Products#

• search-products discovers new products from the web by filters, adds them to a database, and automatically deep-crawls each one to extract features.
• list-products lists every product in a database.
• get-product reads one product with all of its values and description.
• add-product adds a product by name, optionally with a description.
• crawl-website crawls a website into a new product, or into an existing product when you pass its id.
• update-product replaces a product's description document.
• delete-products deletes one or more products from a database.

Attributes#

• list-attributes lists the custom attribute (column) definitions for patents or products.
• create-attribute defines a typed custom column on patent or product records.
• update-attribute renames a custom attribute or replaces its options and config.
• delete-attribute deletes a custom attribute and every value stored for it.
• set-attribute sets or clears one attribute value on a single record.

Mappings#

• list-mappings lists every patent-to-product mapping in a database.
• get-mapping reads one mapping's full per-claim, element-level detail.
• create-mapping creates a blank claim-by-claim mapping skeleton to fill in by hand.
• update-mapping sets the match status and evidence text of a claim's elements.
• delete-mapping deletes one claim from a mapping (or the whole mapping if it was the last).
• generate-mappings generates AI-scored mappings in bulk for products (uses credits).

Pages#

• list-pages lists a database's pages (tables, notes, and dashboards).
• create-page creates a table, note, or dashboard page.
• update-page renames a page.
• delete-page deletes a page (a dashboard's charts are deleted with it).

Tables#

• get-table reads a table's rows and, for a saved page, its view.
• get-column fetches one column's cell values for a table, paginated.
• update-table configures a table page's columns, filters, sorting, and search.

Notes#

• get-note reads a note page's content as a ProseMirror document.
• update-note replaces a note page's content with a new document.

Charts#

• create-chart adds a chart to a dashboard page, optionally sizing it in the 12-column grid.
• list-charts lists a dashboard's charts with their ids, config, and grid layout.
• update-chart replaces a chart's full configuration.
• resize-chart changes a chart's width and height in the grid.
• move-chart reorders a chart within its dashboard.
• delete-chart deletes a chart from a dashboard.

Background jobs#

• get-status lists a database's background jobs so you can poll async work (website crawls, mapping runs, patent imports).

Usage and billing#

• get-usage reports your organization's used and remaining credit and links to top up credit or upgrade the plan.

Reference#

Two tools return guidance rather than data. An agent can read them before doing substantial work to ground itself in how Patalyze is modeled.

• explain-data-model returns a plain-text guide to the data model behind these tools.
• explain-document-format returns a plain-text guide to authoring note and product-description documents.

Connect any MCP client#

Any client that supports remote MCP servers over HTTP with OAuth can connect. Point it at a server URL above; adding the URL starts the OAuth flow, where you sign in to Patalyze and approve access.

Many clients accept a JSON config like this:

{
  "mcpServers": {
    "global-patents": {
      "type": "http",
      "url": "https://data.patalyze.com/mcp"
    },
    "patalyze": {
      "type": "http",
      "url": "https://api.patalyze.com/mcp"
    }
  }
}