O’Reilly MCP Guide

What is MCP?

MCP (Model Context Protocol) is a standard way for AI assistants to securely connect to trusted sources of information in real time. AI tools like Claude, ChatGPT, Microsoft Copilot, and Gemini are powerful, but their output is only as good as the sources they can reach. O’Reilly MCP servers act as a bridge between those AI tools and O’Reilly’s repository of practitioner knowledge.

What we offer

Search MCP

Available to all O’Reilly members

The Search MCP searches O’Reilly and returns relevant results directly, so you can ask your AI tool for learning resources and get direct links to trusted O’Reilly content recommendations. It’s perfect for finding content by topic, format, and level, and curating learning plans based on your needs.

Server URL: https://api.oreilly.com/api/search/v1/mcp

Expert MCP

Enterprise add-on

The Expert MCP pulls context directly from O’Reilly content and cites its sources, so you and your teams get grounded knowledge you can trace to named experts. When your AI tools can also reach your internal systems, they can produce decision-ready analysis backed by O’Reilly expertise.

Server URL: https://api.oreilly.com/api/expert/v1/mcp

How it works

Search MCP

  1. Ask your AI tool for learning resources, for example, “Find me O’Reilly courses on cloud security.”
  2. The AI tool recognizes it has access to the Search MCP server and sends a search request.
  3. The MCP server searches the O’Reilly content library and returns a list of relevant results, including titles, formats, and links.
  4. The AI tool presents the recommendations with direct links to O’Reilly content.

Expert MCP

Includes Search MCP capabilities

  1. Pull vetted expertise from named practitioners directly into your AI tools while you work, for example: “What would an expert principal engineer challenge about this migration plan?”
  2. The AI tool recognizes it can reach the Expert MCP and routes the question to the right tool.
  3. The MCP server analyzes O’Reilly’s practitioner-authored content and returns a synthesized response drawn from it.
  4. The response includes references to specific authors and titles paired with direct links to the exact O’Reilly sources.

AI tools handle and display results in slightly different ways. If results aren’t appearing as expected, rephrase your request to be more specific.

Getting set up

Setup varies slightly by AI tool. Both OAuth (preferred) and token authentication are supported. If your organization sets up the MCP server at the organization level, individual users still need to select the O’Reilly connector in their AI tools and click Connect.

If your organization has purchased Expert Intelligence, be sure to use the Expert MCP server URL.

For individuals

Before you start, make sure your O’Reilly account is active. If you received a welcome email from O’Reilly, click Activate and create your password. You’ll sign in with these credentials when you connect.

Claude Desktop or claude.ai

  1. Go to Settings and select Connectors.
  2. Find O’Reilly in the list your organization has enabled and click Connect.
  3. Your browser will open. Sign in to O’Reilly and click Allow Access.

Claude Code

Then start Claude Code and run the /mcp command. Select the oreilly server and choose Authenticate to sign in through your browser.

ChatGPT

  1. Go to Plugins and find O’Reilly in the list your organization has published.
  2. Click Connect.
  3. Your browser will open. Sign in to O’Reilly and click Allow Access.

For full developer documentation, see the O’Reilly MCP server documentation.

For admins

You’ll add the O’Reilly MCP server to your organization’s AI tools, verify it works, and then let your teams know. When naming the connector, we recommend using “O’Reilly” or “O’Reilly Expert Intelligence” depending on your use case.

OAuth (preferred method)

In general you’ll follow these steps.

  1. Provide the MCP server URL.
  2. If requested, select Streamable HTTP as the transport.
  3. Choose OAuth/OAuth 2.1 as the authentication method.
  4. Accept any terms.
  5. Click Connect. This starts a sign-in to your O’Reilly account. Once signed in, click Allow Access to let the AI tool access your O’Reilly account.

Note: Static client credentials are not currently supported.

Claude (organization setup)

  1. Open Claude Desktop or claude.ai.
  2. Go to Organization settings and click Connectors in the left navigation.
  3. Click +Add and choose Custom → Web.
  4. Enter a connector name such as “O’Reilly” and the MCP server URL.
  5. Click Add.
    Claude organization Connectors settings showing the Custom Web connector configuration for O’Reilly
  6. If you aren’t prompted to connect, click the three vertical dots and select See details.
  7. Click Connect under Tool permissions and you’ll be redirected to your browser to sign in to O’Reilly and allow access.
  8. Your employees will now see O’Reilly in Claude under Settings → Connectors, and they’ll need to click Connect to authenticate and begin using the MCP server.

ChatGPT (organization setup)

  1. Open ChatGPT Enterprise.
  2. Go to Apps and click +Create.
  3. Enter a connector name such as “O’Reilly” and the MCP server URL under Connection.
  4. Choose OAuth as the authentication method and leave the advanced OAuth settings as they are. Click I understand for each of the notices and click Create.
    ChatGPT Enterprise Apps configuration showing the O’Reilly connector with OAuth authentication settings
  5. You may be directed to your browser to sign in to O’Reilly and allow access.
  6. Click Publish.
  7. Go to Plugins and confirm the app is available. You can also optionally configure which roles have access.
  8. Your employees will now see O’Reilly in ChatGPT under Plugins, and they’ll need to click Connect to authenticate and begin using the MCP server.

Before you announce

Connect your own account and run the checks in “Check That It’s Working” below. Once you’ve confirmed the connection, share the individual setup prompt from the top of this guide with your teams. It’s the fastest path from your announcement to their first answer.

Token access (alternate method)

If your AI tool does not support OAuth, you may be able to use token authentication instead. Treat tokens like passwords: store them in your tool’s configuration or a secrets manager, and never paste them into a chat conversation.

Individual user access

  1. Sign in to O’Reilly.
  2. Go to the MCP Tokens page.
  3. Give the token a name and set an expiration date.
  4. Click Generate token.
  5. Save your token immediately in a secure location. It won’t be displayed again.
  6. Use the MCP server documentation to complete setup.

From the MCP Tokens page, you can view and revoke any tokens you have created.

Admins: Organization-wide access

  1. Log in to the Admin Console.
  2. Navigate to Integrations → API Tokens.
  3. Click Create token.
  4. Select Content MCP as the token type.
  5. Give the token a name and set an expiration date.
  6. Click Continue—your token will be generated.
  7. Save your token immediately in a secure location. It won’t be displayed again.
  8. Use the MCP server documentation to complete setup.

From the API Tokens page, you can view and revoke any tokens you have created.

Check that it’s working

Test the Search MCP. Ask your AI tool: “Find me three O’Reilly courses on cloud security.” You should get course titles with direct links to learning.oreilly.com.

Test the Expert MCP (if your organization has the add-on). Ask: “What do experts recommend for setting reliability targets on a brand-new service? Cite sources from O’Reilly.” You should get a synthesized answer that names specific O’Reilly authors and titles, with links to each source.

Use cases and examples

Search MCP

Find the right content fast

Find me O’Reilly courses on Kubernetes for someone who knows Docker but hasn’t run a production cluster. Include one book and one video course.

Build a learning plan

I’m a backend engineer preparing for the AWS Solutions Architect Associate exam in eight weeks. Build me a week-by-week study plan from O’Reilly content, mixing courses, practice material, and one book.

Keep a team current

Find upcoming O’Reilly live events on GenAI for engineering teams in the next two months, plus two self-paced alternatives for people who can’t attend live.

Expert MCP

These prompts work in any AI tool connected to the Expert MCP. Where a prompt references your own documents or systems, attach the document or connect the system first.

Tip: No matter your use case, a specific, decision-shaped prompt gets the best results.

Onboard into a new domain

Use case: You inherited a data engineering team but have never managed one, and you need a 30-60-90-day plan.

Build me a 30-60-90-day plan to ramp into managing a data engineering team. I have a background in backend services but limited data infrastructure experience. Use O’Reilly’s data engineering and management content to scope what I should learn, what conversations I should have with my team, and what early decisions I should avoid.

What you’ll get back: A staged ramp plan grounded in sources like Joe Reis and Matt Housley’s Fundamentals of Data Engineering and Camille Fournier’s The Manager’s Path, with links to the exact chapters.

Evaluate team structure

Use case: Your platform team hit 14 engineers, and velocity is slowing. You need to decide whether to split, restructure, or change the inflow of work.

Our platform team has grown to 14 engineers and velocity is slowing. Use O’Reilly to identify the team scaling patterns and failure modes that apply at this size, and what interventions do expert frameworks recommend: Split it, restructure, or change how work flows in?

What you’ll get back: Named patterns and interventions drawn from sources like Matthew Skelton and Manuel Pais’s Team Topologies, Camille Fournier and Ian Nowland’s Platform Engineering, and Morgan Evans’s The Engineering Manager’s Handbook.

Define SLOs for a new service

Use case: You’re launching a new payments service with no production history and need reliability targets that product and customer teams will accept.

We’re launching a new payments service. Draft an SLO framework that covers error budgets, how to handle the rollout period with no historical data, and how to negotiate these targets with product and customer teams. Cite expert sources from O’Reilly.

What you’ll get back: A framework built on sources like Alex Hidalgo’s Implementing Service Level Objectives, including the chapters on setting targets without history and getting stakeholder buy-in, alongside Site Reliability Engineering and The Site Reliability Workbook.

Estimate project timing

Use case: You need to know how long a project will really take before the team spends weeks scoping it.

Given this backlog and design doc, estimate the realistic duration and the biggest schedule risks, grounded in expert delivery frameworks. Cite sources from O’Reilly.

What you’ll get back: An estimate and risk register grounded in sources like Steve McConnell’s Rapid Development, Tom DeMarco and Timothy Lister’s Waltzing with Bears, and Tom Kendrick’s Identifying and Managing Project Risk.

Review an RFC

Use case: You’re reviewing an RFC for adopting a service mesh and need to know what it handles well, what it underestimates, and where to push back before it moves forward.

Review the attached RFC for adopting a service mesh. Identify what it handles well, what it underestimates, and what questions I should push back on. Pull from O’Reilly’s content on service mesh adoption and trade-offs, and cite sources.

What you’ll get back: A structured review citing vetted content from independent practitioners, like Istio: Up and Running by Lee Calcote and Zack Butcher, rather than vendor blogs with products to sell.

Build a role-based learning program

Use case: You rolled out AI coding tools to all of engineering and adoption is uneven. You need a program that meets each level where it actually is.

We rolled out AI coding agents to all of engineering, and adoption is uneven. Use O’Reilly to build a role-based learning path: What capabilities should senior engineers, mid-level engineers, and new hires each build first to use these tools well, and what should the first 90 days of the program look like? Cite expert sources.

What you’ll get back: A tiered program grounded in an analysis of expert content on the O’Reilly learning platform, with each recommendation for senior engineers, mid-level engineers, and new hires traced to a named source.

Evaluate a strategic shift

Use case: You’re deciding whether to adopt platform engineering across a 200-engineer org and need a memo for leadership.

I’m drafting a memo on whether we should adopt platform engineering practices and internal developer platforms across our 200-engineer org. Use O’Reilly to synthesize the current expert consensus. Where do practitioners agree, where do they disagree, and what are the most common failure modes at our scale? Cite sources.

What you’ll get back: A leadership-ready synthesis that maps where practitioners agree, where they disagree, and the failure modes most common at your scale, with each point traced to named O’Reilly authors and titles you can cite in the memo.

Cross-system research on a team or feature

Use case: You want to know how a specific team is doing: what they’ve shipped, the incident load, and the strategic risks. This one shines when your AI tool is also connected to your internal systems.

Pull together a picture of the recommendations team. What have they shipped this quarter from Jira and GitHub? What does their incident load look like? What strategic risks are showing up in their retros and design docs? Then evaluate where they are, using frameworks from O’Reilly. Cite sources and flag the most important issues for me to act on this week.

What you’ll get back: A prioritized read on the team that combines your own system data with expert frameworks from the O’Reilly learning platform, with the most important issues flagged for the week ahead and each recommendation traced to a named source.

Agent skills

Coming soon for Expert Intelligence customers

Expert Intelligence will include expert-built agent skills that package proven methods, like architecture review, migration planning, and more, so the Expert MCP is called at the right moment. We’ll add setup instructions here as these become available.

Troubleshooting

I don’t see O’Reilly in my AI tool. Your administrator may still be rolling it out. Check under Settings → Connectors (Claude) or Plugins (ChatGPT). If it’s not there, ask your administrator.

I can’t sign in when I click Connect. You need an active O’Reilly account. If you received a welcome email from O’Reilly, click Activate and create your password, then try connecting again.

I’m connected, but answers don’t cite O’Reilly sources. First, rephrase your prompt to be more specific and ask for cited sources, or mention O’Reilly directly. If that doesn’t help and your organization has the Expert MCP add-on, disconnect and reconnect the O’Reilly connector in your AI tool.

Results aren’t appearing as expected. Different AI tools handle and display results in different ways. Try rephrasing your request to be more specific about topic, format, or level.

A colleague has access and I don’t. Access follows your O’Reilly account. Make sure your account is activated and you’ve clicked Connect in your AI tool. If it still doesn’t work, ask your administrator.

Where do I put my token? In your AI tool’s configuration or a secrets manager. Never paste a token into a chat conversation, and revoke any token you’ve exposed from the MCP Tokens page.

Data privacy

The O’Reilly MCP server only accesses the information needed to search for and return content recommendations. It does not store or retain your employees’ conversations with their AI tools.

However, your organization should be aware that the AI tool decides what context is necessary to send to the O’Reilly MCP server to receive the best response. O’Reilly cannot control what information the AI tool sends to the MCP server. We recommend reviewing the data retention and privacy settings of any AI tools your organization connects to the O’Reilly MCP server and establishing internal guidelines around what information your employees share with AI tools.