mcpField
P
PatentReply

PatentReply

Remote
patentreply.aiLegal & Government0Updated 2d ago

About PatentReply

PatentReply is a hosted MCP service for reviewing and working with public U.S. patent prosecution records in AI clients. It organizes claims, Office Actions, references, and specifications, with cited evidence and practitioner approval required before response drafting.

Key features

  1. 01Organized USPTO prosecution records
  2. 02Claim and Office Action review
  3. 03Cited source passages
  4. 04MPEP and legal-authority retrieval
  5. 05Practitioner-approved response drafting

Use cases

  1. 01Assess examiner rejections and citations
  2. 02Prepare examiner interview agendas
  3. 03Develop amendment and response plans
  4. 04Check a draft against the prosecution record

MCP support

The hosted Streamable HTTP MCP server gives AI clients access to patent records, legal-authority tools, and case workflows. It uses OAuth sign-in; response drafting tools require registered practitioner approval of the plan.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.patentreply.ai/mcp/v2
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

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

Details

From the project README

PatentReply MCP — USPTO Patent Prosecution for Claude and ChatGPT

Connect your AI to the patent file wrapper. Open a U.S. patent matter in Claude, ChatGPT, or Cursor and work with an assistant that has the prosecution record organized and indexed: claims, Office Actions, cited references, and the specification.

Review a rejection, check the examiner's citations, prepare for an interview, or develop a response plan. PatentReply supplies the record and workflow; you direct the strategy and approve the plan before response drafting.

Connect PatentReply · Follow the walkthrough · Documentation

Start free: one new public matter per month, no credit card. See pricing and plan features.

How it works

  1. Connect once. Add the hosted Model Context Protocol (MCP) server to your AI client and approve the PatentReply sign-in.
  2. Open a matter. Give your AI an application number. PatentReply retrieves the public USPTO file wrapper and builds the working record. The first compilation can take a few minutes.
  3. Work through the case. Ask questions, inspect the source passages, and develop a response. An attorney approves the response plan on PatentReply before the gated drafting tools generate the response and claim listing.

After connecting, replace the bracketed text with your application number:

Open U.S. application [your application number] in PatentReply. Summarize the current claims and prosecution posture, and flag any missing or disputed record material before we discuss strategy.

What you can do

TaskTry asking your AI
Understand the prosecution history"What changed in this application, and which claims are currently pending?"
Analyze an Office Action"Review each rejection and show the claims and evidence the examiner relies on."
Verify a cited passage"Show the cited paragraph of the primary reference and compare it with claim 1."
Check patent prosecution authority"Retrieve MPEP 2143.01 and explain how it relates to this rejection."
Prepare an examiner interview"Help me prepare an interview agenda and fallback positions for this matter."
Plan amendments and a response"Develop response options for each rejection, with specification support for proposed amendments."
Review a draft before filing"Give me an upload link for my response draft, then run QC against the record."

Explore the workflows and tools and walk through an Office Action review. Feature access and usage allowances depend on your plan.

Connect your AI

ConnectionDetails
Hosted MCP serverhttps://mcp.patentreply.ai/mcp/v2
TransportStreamable HTTP
AuthenticationOAuth sign-in; no API key
ClientGet started
Claude web, desktop, or mobileAdd a custom connector using the server URL, then approve the sign-in. Claude guide.
ChatGPTAdd PatentReply as a custom MCP plugin with OAuth. ChatGPT setup and compatibility URL.
Claude CodeInstall the plugin below, or use the MCP-only connection.
CursorAdd mcp/cursor.mcp.json to your MCP configuration. Cursor guide.
Other clientsUse a client that supports Streamable HTTP and PatentReply's OAuth flow. Connection details.

The connection guide covers authentication, first use, and troubleshooting.

Claude Code plugin

Install the packaged skills, commands, and MCP connection together:

claude plugin marketplace add patentreply/patentreply-mcp
claude plugin install patentreply@patentreply

Start a Claude Code session, run /mcp, and complete the PatentReply sign-in. Then try /patentreply:status with an application number, or use ordinary language. Other commands are /patentreply:oa, /patentreply:respond, and /patentreply:interview.

The plugin already declares the hosted MCP server as patent-kb. Choose this installation or the MCP-only connection; see existing connections if you already added the endpoint manually.

Evidence and attorney control

PatentReply makes the compiled claims, Office Actions, cited references, specification, and prosecution history available to your AI. Record citations use pkb:// verification addresses into the stored matter. Legal-authority tools retrieve MPEP and other supported authority text for the analysis.

The workflow checks the record before drafting and surfaces missing source material, unresolved citations, and disputed claim text. A source address lets you inspect the evidence; the attorney still evaluates the conclusion.

Response planning and drafting are separate stages. A plan revision is submitted for a registered practitioner's approval on the PatentReply approval page. After approval, the hosted service generates the response Word document and the 37 CFR 1.121 claim listing for attorney review. A confirmation in chat does not sign the plan, and PatentReply does not file papers with the USPTO.

Public repository and service

This is the official public repository for PatentReply setup instructions, registry metadata, and the Claude plugin. The hosted service is operated by Patent Agentic LLC.

ResourceLocation
Connection instructionsdocs/connect.md
Workflows, skills, and tool overviewdocs/workflows.md
Office Action review walkthroughexamples/office-action-review.md
Pricing, records, and privacy questionsdocs/faq.md
Claude plugin contentsplugin/
MCP Registry metadataserver.json, server name ai.patentreply/patentreply
Glama ownership metadataglama.json
Maintenance and discovery checklistdocs/maintainers.md

The plugin/ tree is generated from the PatentReply build and replaced on release. Report changes through an issue so they can be made in the source build. See maintenance guidance.

Privacy and support

PatentReply receives the arguments your AI sends in its tool calls, including any text or documents you choose to supply through the workflow. It does not independently read your conversation history, client memory, or local files. PatentReply does not use your matters to train models. Your AI provider's handling of the conversation is governed separately.

Privacy · Security · Terms · Pricing

For connection problems or documentation feedback, open an issue or email [email protected]. Include the client and the error message; send confidential matter information through the appropriate private support channel.

PatentReply is software, not a law firm. The registered practitioner remains responsible for every filing. See the FAQ for scope and limits.

License

The files outside plugin/ are released under the MIT License; see LICENSE for its scope. The generated plugin/ tree is proprietary as stated in its manifest. The hosted service is governed by its Terms.

M8ven Score

Comments

Markdown supported

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.