MCP Server

GoVeda provides an MCP server that lets AI assistants call the same patent tools available in the REST API. If you use Claude Code, Claude Desktop, ChatGPT, or any other MCP-compatible client, you can search patents, retrieve patent content, and generate reports directly from your AI workflow.

The MCP server uses your GoVeda account credits. Check your balance with the get_usage tool or on the Usage page.

Connect your AI assistant

The MCP server endpoint is:


https://mcp.goveda.com/mcp

Authentication uses OAuth. Your AI client opens a browser window for you to sign in with your GoVeda account the first time you connect.

Claude Code

Run this command in your terminal:


claude mcp add goveda-patent --transport streamable-http https://mcp.goveda.com/mcp

Claude Code handles OAuth automatically. The first time you use a GoVeda tool, a browser window opens for sign-in.

Claude Desktop

Add this to your Claude Desktop MCP config file (claude_desktop_config.json):


{
  "mcpServers": {
    "goveda-patent": {
      "type": "streamable-http",
      "url": "https://mcp.goveda.com/mcp"
    }
  }
}

Restart Claude Desktop. The first tool call will prompt you to sign in.

Other clients

Point your MCP client to the endpoint above using Streamable HTTP transport. The server supports standard OAuth 2.0 discovery at https://mcp.goveda.com/.well-known/oauth-authorization-server.

You can also authenticate with an API key by passing the X-API-Key header if your client supports custom headers.

Available tools

The MCP server exposes eleven tools. Your AI assistant discovers them automatically once connected.

Search & discovery

Tool	What it does	Credits
`semantic_patent_search`	Semantic search across 220M+ patents by topic or invention description. Returns ranked results directly.	~50 (10 results) to ~320 (100 results)
`prior_art_search`	Finds prior art for a given patent. Returns a `search_id` to poll with `get_search_status`.	~690 (10 results)
`get_search_status`	Checks the status of a search or prior art request.	Free

Patent content

Tool	What it does	Credits
`get_patent`	Retrieves content for a single patent. You can request specific sections (abstract, claims, description, dates, parties, classifications, legal status, citations, family) or all of them.	1
`batch_get_patents`	Same as `get_patent` but for multiple patents in one call.	1 per patent

Reports

Tool	What it does	Credits
`generate_novelty_report`	Runs a novelty and patentability assessment for an invention description. Returns a `report_id` to poll with status tools.	720
`get_report_status`	Checks report generation status.	Free
`get_report_summary`	Returns the executive summary, verdict, threats, and a lightweight patent directory from a completed report.	Free
`get_report_patent_analysis`	Returns detailed per-patent analysis (relevance, novelty features, blocking potential, key differences) for specific patents from the directory.	Free

Account

Tool	What it does	Credits
`get_usage`	Returns your remaining credit balance and billing period.	Free
`start_free_trial`	Activates a 14-day free trial with 10,000 credits. Only available for new users.	Free

How async tools work

semantic_patent_search polls internally and returns final results in a single call (typically 5-15 seconds).

prior_art_search and generate_novelty_report return immediately with an ID. Your AI assistant needs to call the corresponding status tool to check progress:

Submit the request

Call prior_art_search or generate_novelty_report. You get back a search_id or report_id.

Poll for results

Call get_search_status or get_report_status with that ID. If the status is still in_progress, wait and try again.

Read the results

When the status is completed, call get_report_summary to get the executive summary and patent directory. Use get_report_patent_analysis to drill into specific patents. For prior art searches, the results are included directly in get_search_status.

Most AI assistants handle this polling loop automatically.

Credit costs

Credit costs for MCP tools are the same as the REST API. Search costs scale with the number of results you request.

Operation	Cost
Semantic search (10 results)	~50
Semantic search (100 results)	~320
Prior art search (10 results)	~690
Patent content	1 per patent
Generate report	720
Status checks, usage	Free

Prior art searches and reports are expensive. Your AI assistant should confirm with you before running these operations.

Example conversation

Here is what a typical interaction looks like once connected:

You: Search for patents about solid-state battery electrolytes using sulfide glass Assistant: (calls semantic_patent_search with your query, returns 10 results with titles, UCIDs, and relevance scores)

You: Get the full claims for the top result Assistant: (calls get_patent with the UCID and sections=claims)

You: How many credits do I have left? Assistant: (calls get_usage, reports your balance)

Differences from the REST API

The MCP server wraps the same backend as the REST API, with a few differences:

Authentication. MCP uses OAuth (browser sign-in) instead of API keys, though API keys also work.
Transport. MCP uses Streamable HTTP, not standard REST.
Patent content. The description and claims fields are truncated to 8,000 characters per section for compatibility with AI context windows. The REST API returns the full text.
Search behavior. semantic_patent_search polls internally and returns final results. The REST API returns a search_id for you to poll yourself.
Report retrieval. MCP splits report access into get_report_summary (overview + patent directory) and get_report_patent_analysis (detailed per-patent analysis). The REST API returns the full report in one call.

Troubleshooting

“Insufficient credits.” Check your balance with get_usage and add credits on the Plans & Billing page. New users can activate a 14-day free trial with 10,000 credits by calling start_free_trial.

“FREE_TRIAL_AVAILABLE”. Your AI assistant should call start_free_trial to activate your free trial, then retry the original request.

OAuth sign-in loop. Clear your browser cookies for mcp.goveda.com and try connecting again.

Tool not found. Make sure your MCP client is configured with the correct endpoint (https://mcp.goveda.com/mcp) and Streamable HTTP transport.