Outreach MCP Server Overview

Objective

Learn about Outreach MCP Server and how it allows other AI Agents to utilize Outreach knowledge directly inside their workflows. In a server capacity, Outreach acts as a producer (server) of knowledge and a facilitator of actions for MCP Clients. To learn more about Outreach as an MCP Client, refer to the MCP Client documentation.

Outreach MCP Server is built to support Anthropic’s Model Context Protocol (MCP) standard and can be utilized by any MCP Client that follows the MCP specifications. Any user who wants to utilize Outreach MCP Server must be an active, licensed seat in an Outreach instance and have the Amplify add-on package enabled in their plan.

Applies To

• Outreach Admins who want to set up Outreach via MCP for their organization
• Outreach Users who want to use Outreach via MCP in their daily workflow tools like Claude, ChatGPT, Microsoft Copilot, Google Gemini, and so on
  
  

Before You Begin

Outreach MCP Server requires the following for users trying to access Outreach via MCP:

• The organization must have Amplify enabled with active credits.
• The user must be active and have access to at least one Outreach instance.
• MCP Clients like Claude, ChatGPT, etc. might require additional licenses and environmental setups.

Note: If you do not have the Amplify add-on, or you are unsure, reach out to your AE to inquire about it.

Overview

What is MCP?

AI applications are only as useful as the knowledge they can generate from the data they have access to. The real world is a landscape of disconnected data islands, hosted in their own servers and controlled by the hosting product. For useful and usable experiences, AI applications need to connect to these data sources. This is where MCP, or Model Context Protocol (MCP), comes in.

MCP is an open-source standard from Anthropic for connecting AI applications to external systems for tools — knowledge, insights, and actions that the local system doesn’t have. MCP standardizes how LLMs communicate with each other. Think of MCP as context in motion — your Outreach AI can share or draw to and from the broader ecosystem of tools your company already uses, and those systems can talk back to Outreach.

You can learn more about MCP from Anthropic’s official documentation. You can install Outreach's Official MCP Server directly in Claude.

Outreach MCP Server

In a server capacity, Outreach is a producer of knowledge and facilitator of actions for MCP Clients. Other AI Agents, or end users, can now utilize Outreach’s data and draw best-in-class insights directly inside their revenue workflows.

Outreach MCP Server is available to all Outreach customers to connect to. While it’s an authenticated, open server, it is expected to be most useful for MCP Clients that integrate Outreach value into their workflows and end-user experiences.

Outreach envisions that customers can use Outreach-exposed MCP Tools in all sorts of applications, including LOB tools, open MCP Clients like Claude, OpenAI, and Slack, and chat applications like Agentforce or Copilot.

Note: We’re constantly adding more tools through collaboration with our customers and creating innovative scenarios with partners. Please make sure to refresh tools time to time.

Setup

Enabling Outreach MCP Server

Note: If you do not see the options below available to you, you may not have the Amplify add-on. Reach out to your AE to inquire about the Amplify add-on product.

Before MCP Clients can connect to it, Outreach MCP Server must be enabled. This setting is per organization and only exposes the data of that specific instance. This action is only available to the organization's admins.

To enable the MCP Server:

1. Log in to Outreach as an Admin.
2. Click Administration (settings) > Organization > Org Info.
3. Navigate to Gen AI.
4. Toggle on MCP Server.
5. There are two optional settings which allows you to control whether you want to allow create (ON by default) and delete (OFF by default) actions over MCP.

Authentication

Outreach enforces that only authenticated, authorized, and licensed users can connect to the server. We also require that the instance, and the user, has Amplify enabled. For supported features, Outreach follows MCP’s Authorization standards as published on November 11, 2025.

To access Outreach tools, the user must sign in with their Outreach credentials and provide any additional factors required by their organization. The MCP Client will be set up using the credentials you provide, so make sure you are providing the credentials of the instance you want to use.

Using Outreach with MCP Clients

For the purposes of this documentation, we’ll use Claude as an example, but any MCP Client can connect to Outreach Server and utilize its knowledge and insights, as long as the user is able to authenticate with their credentials against Outreach and has successfully fulfilled the prerequisites. Outreach also requires that the client itself support OAuth 2.1 user authentication and Dynamic Client Registration.

Option 1: Browse connectors (RECOMMENDED)

1. Our recommendation is that you install the official Outreach connector. In Claude, go to Customize > Connectors > +.
   
2. In the Connectors dialog, search for Outreach, then click Connect
   
3. Claude should pick up Outreach MCP Server automatically. Complete authentication with Outreach if you are not signed in.

Option 2: Add a custom connector
This is not the recommended approach for daily users. This option is available to Admins or Developers who want to connect to non-production instances of Outreach (like Sandbox, or Testing instances)

1. Log in to Claude
2. Click Customize > Connectors > Hit the + button > ... Add custom connector
   
3. Enter the following information:
   • Name – Outreach (or your preferred name)
   • Server URL – https://api.outreach.io/mcp
   • You can ignore the Advanced settings
     
4. Click Add
5. Next to Outreach, click Connect.
6. Sign in with your Outreach credentials in the browser window. If you have multiple instances, make sure you’re using the credentials for the organization whose MCP Server you want to connect to.
7. Complete authentication by providing multi-factor authentication if required by your organization.

Completing the connection

If required, Outreach will prompt you to consent to the information that will be shared with Claude (or another MCP client if you’re not using Claude). Once you have authenticated and provided consent, Claude will be connected to Outreach MCP Server and ready for you to run queries.

If you want to view the tools list or configure individual tools, you can do so from Customize > Connectors > Outreach > Configure.

Note: Outreach recommends that you keep all tools enabled and provide “Always allow” permission for the most productive experience.

MCP Client Availability

Any AI application that supports the MCP protocol, user-level auth, and Dynamic Client Registration should be able to connect to Outreach MCP Server. Here are the instructions to connect the most popular clients:

• Claude – Available. Instructions are in this document above.
• ChatGPT – Available for both ChatGPT and Codex
• Microsoft Copilot – Available. You can use it with Microsoft Copilot and Microsoft 365 Apps.
• Gemini – Available. Install through CLI.
• AWS Bedrock – Available. You must be signed in to view.
• Development tooling for major IDEs (like Visual Studio Code) – Available. Install through CLI or Config.

Our up-to-date listing can be found in Outreach Marketplace > MCP Connectors.

Usage and User Experience

MCP is a standalone protocol and can be utilized by any end-user GenAI application that supports the protocol. This opens a whole world of possibility for how best to utilize Outreach’s knowledge inside your daily workflows.

With GenAI, the most common experiences are either a chat-based (user-initiated) interaction or Agent-to-Agent autonomous collaboration. Let’s look at those closely.

Chat Experience

Outreach MCP Server can be connected directly with GenAI (LLM) chat experiences within Claude, Copilot, OpenAI ChatGPT and Codex, etc. or your LOB applications, once you have set up the MCP clients. You can run queries as you normally would as part of your daily revenue workflows, and the chat should connect automatically to Outreach MCP Server and display insights and knowledge from Outreach.

Outreach MCP Server comes with a robust set of tools. Here are some sample queries you can run:

Prepare for meetings: “Help me prepare for my next call with Acme”

Create content referencing conversations: “Generate a best practice guide for each feature discussed in the last meeting with Acme”

Analyze what’s working and why: “What sequences from our customer conference have the best reactions, and what are they?”

Plan smarter objection-handling: “Tell me Contoso’s concerns around our solution and how I should combat that?”

Note: The quality of the response will be determined by the richness of data available to Outreach.

Agent-to-Agent Autonomous Collaboration

If you are running Agentic AI that is fully autonomous (for example, task completions, automatic workflow execution, etc.), these agents can also utilize insights and knowledge from Outreach MCP Server as part of their flow.

From an MCP point of view, these are fundamentally the same client-server communication as a chat-based interaction. The key difference is that, in this case, the agent runs fully autonomously.

Note: Outreach does not support Client Auth-only clients. For security reasons, Outreach enforces that an active user credential, SSO or non-SSO, must be provided over MCP, and the Profile assigned to that user account will be applied. If you plan to use an Agentic Identity, that identity must be an active, licensed user in Outreach.

Outreach MCP Tools

Pro Tip: If you want a detailed rundown on the latest tools, descriptions, or sample queries with request-response, you can always ask Claude, ChatGPT, or another LLM you’ve connected to Outreach MCP Server. You can even ask it to create shareable materials to send around to your teams.

Outreach exposes the following tools:

• Retrieve transcripts and recordings from Kaia calls – kaia_meeting_search
• Retrieve email content – emails_search
• Lookup existing sequences – sequence_search
• Track prospect progress through a sequence – sequence_state_search
• Add prospects to a sequence – sequence_add_prospects

• Lookup prospect by ID – prospect_get_by_id
• Search prospects by name, account, opportunity, or owner – prospect_search
• Lookup prospect by external IDs like CRM IDs – prospect_search_by_external_id
• Create a new prospect – prospect_create

• Lookup account by account ID – account_get_by_id
• Search accounts by name, owner, or team – account_search
• Lookup account by external IDs like CRM IDs – account_search_by_external_id
• Ask general QnA and ask for Account insights – account_answer_question
• Create a new account – account_create

• Lookup opportunity by ID – opportunity_get_by_id
• Search opportunities by name, account, owner, or date range – opportunity_search
• Lookup opportunity by external IDs like CRM IDs – opportunity_search_by_external_id
• Ask general QnA and ask for Opportunity insights – opportunity_answer_question
• Create a new opportunity – opportunity_create
• Retrieve opportunity stages (e.g. Discovery, Negotiation, Closed Won) – opportunity_stage_fetch

• Lookup user by ID – user_get_by_id
• Search users by name or email – user_search
• Get the current authenticated user – current_user

• Retrieve scheduled calendar events and meetings – calendar_events_fetch

• Fetch filterable and custom fields for prospects, accounts, or opportunities – filter_fields_fetch
• Fetch available input fields before creating a record – input_fields_fetch
• Retrieve prospect lifecycle stages configured for your org – stage_fetch

Exact configuration of the tools and sample responses can be found in our developer documentation for MCP Server. At this moment, Outreach has not exposed any resources or prompts.

Data Security

Generally speaking, you can treat LLMs as another API endpoint, or client application, and secure the endpoint just as you would any other API endpoint or client application. MCP uses an HTTP endpoint, so you can use your web postures as a starting point. You can also follow general IAM best practices as they apply to OAuth 2.1. Some of the patterns are given below.

Post-Auth Data Protections

Before sharing any data over MCP tools, Outreach enforces that the user must sign in, have an active user account, and have a valid license assigned. Outreach also enforces any Outreach (RBAC) Profile assigned to the user, so users can only access data allowed through their Profile permissions, the same as accessing Outreach through our web experience.

Pre-Auth Access Restrictions

If you want to enforce restrictions to access Outreach MCP Server before a user has authenticated, Outreach has no control over that. Our recommendation instead is to follow your Identity Provider’s (IDP’s) best practices. If you want to enforce endpoint management, use MDM and MAM tools. You can also apply any vendor-specific policies.

Here are some of the things you could consider:

Use your IDP tools – Most IDPs allow you to restrict access to a specific URL and control access to specific clients, users, or employee groups.

Use your MDM/MAM tools – If you want to manage the client access endpoints, enforce data or account security policies, or enforce proxy restrictions, MDM/MAM tools will help you lock it down.

Use AI vendor recommendations – Most popular AI platforms have their own security tools and guidance you could use. It might require specific licenses. Claude, for example, recommends Enterprise licenses, SSO restrictions, and tenant restrictions you could use to manage Claude.

Note: Outreach MCP Server cannot distinguish between different deployments or change posture. It’s up to you to test and verify that Outreach MCP Server continues working, and that all its tools are usable, after setting up restrictions.

Additional Information

Frequently Asked Questions

What is MCP in simple terms?

MCP (Model Context Protocol) is a universal language that lets AI agents and tools talk to each other across different platforms. It removes silos so systems can share context and take actions seamlessly.

How is MCP different from an API?

APIs are point-to-point connections between two systems. MCP is a shared protocol that enables many AI agents and platforms to collaborate dynamically without custom integrations. Think of MCP as interoperability of AI agents at scale.

Is MCP a product or a standard?

MCP is a protocol (a standard), not a standalone product. Outreach uses MCP to make its AI agents interoperable with other tools your customers already use.

Where can I find MCP clients I can use?

MCP allows you to use any compliant client to connect to servers, and you are welcome to use any well-known MCP Clients. To try out Outreach MCP Server, the fastest way would be to use Claude or ChatGPT. For more technically minded people, or if you don’t want to pay for a client yet, you can use Visual Studio Copilot.

Outreach MCP Server really shines when the knowledge and insights show up in daily revenue flows. If your organization has access to Salesforce Agentforce, Microsoft Copilot, or Google Gemini, they all support connecting to MCP Servers and provide an integrated experience. You can also use tools like Slack or Microsoft Teams if you want to use Outreach directly in your team-selling communication workflows.

Some organizations might restrict usage of public MCP Clients. Your organization admin would be the best group to reach out to. At minimum, your organization will need to enable MCP Server for your Outreach instance before you can use it.

Will MCP work with my existing tools?

Yes, as long as your tool can connect over MCP. MCP is designed for interoperability across platforms. Outreach will support major systems like Salesforce, Microsoft Copilot, and others at launch, with more partners added over time. If a customer’s tools support MCP or have an MCP server/client, they can participate in this ecosystem without custom development.

How does Outreach pick the instance to get me knowledge and insights from?

Outreach uses the credentials you provide to connect to the appropriate Outreach instance, so make sure you’re entering the credentials based on which instance you want to connect to. Most clients open your default browser to authenticate you, so make sure you’re signed in with an appropriate Outreach account on your default browser before you sign in to a client.

Can an MCP client connect to multiple Outreach instances?

One client can only connect to one server per the MCP protocol. However, you can create an MCP host to connect to multiple Outreach MCP servers if the host is able to maintain credentials from multiple accounts. This is not something Outreach controls, and you’ll need to refer to the documentation of your host AI application.

One thing to note however is concurrent connections - when you connect to 1+ Outreach instances to the same LLM session, this could be confusing to the underlying LLM since the LLM can solicit valid response for the same query from multiple Outreach instances. Talk to your LLM vendor on the best way to approach this. Outreach recommends that you connect to one instance at a time. If needed, you are welcome to create multiple browser profiles and connect to one Outreach instance for each profile.

Which version of the MCP spec does Outreach support?

For supported features, Outreach follows MCP’s Authorization standards as published on November 11, 2025.

What is the pricing for MCP?

MCP Server is available to customers with Amplify. Please refer to your pricing guidance or Outreach partner for details on pricing. Once enabled by an admin, it is available to any licensed and active user. API throttle limits apply. If you’re not sure whether you have this, or want to inquire about it, please reach out to your AE.

Why is the Enable MCP Server toggle greyed/grayed out in Org settings?

Make sure you have Amplify enabled for your organization. If you do not have the Amplify add-on package, please inquire about this with your AE if you’re interested in adding it to your plan.

Does MCP support both manual and agentic workflows? Does it support technical or non-user identity?

Yes. MCP connectors work in both scenarios. They support manual workflows, like sellers requesting actions via Ask Outreach, and agentic workflows, where AI agents use external insights, such as ZoomInfo or Crayon, to generate outputs like personalized emails.

MCP supports authentication using non-user identities, as long as a valid authentication token is provided. Outreach does not differentiate between human and non-human identities at the protocol level. However, the identity must still be associated with a licensed user in Outreach. Client credential–based authentication is not currently supported.

Are there any limits to high usage?

Yes, API throttle limits apply. Please check in with your Outreach partner if you are hitting API calling limits and options available to you for increasing the limits.

With MCP, are existing integrations like APIs, SDES, and Snowflake Enrichment still relevant?

MCP is not designed to move large datasets or store third-party insights inside Outreach. Customers still need traditional integrations like SDES and Snowflake Enrichment when they want their reps to see enriched data directly inside Outreach, when they want enrichment to run automatically at scale, and when they want permanent fields populated for reporting or workflows. Customers can continue to use APIs to pull data into their existing workflows.

MCP connections are ideal for letting agents fetch just-in-time context from another system, generate guidance, and take action across platforms. These agent-to-agent interactions do not replace the value of persistent, visible, governed data inside Outreach.

How can MCP access be controlled using roles or profiles?

MCP access follows existing role and profile permissions in Outreach. Data accessed through MCP is governed by the same permissions assigned to each user, making it straightforward to limit access for specific users or groups using standard profile configuration.

If you want exactly the same access control on Server and Client, make sure you have the same permissions applied on both Server and Client.

Can a specific user’s MCP access be revoked?

Yes, but not through Outreach. Access can be revoked by an admin through the organization’s Identity Provider (IDP). Once revoked, the user or system will no longer be able to authenticate or access Outreach via MCP. Please consult your IDP documentation for secure or conditional access.

How many MCP calls are used with one query?

MCP server calls depend on the agent used and how it reasons. Here’s how this works, using Claude as an example:

• User enters a query like “Help me prepare for my next call with Acme” into Claude.
• Claude reasons to understand if it has enough context to respond.
• Claude reasons that it’ll need help from Outreach.
• Claude asks Outreach via MCP:
  • API call 1: provide details on Acme Account.
  • API call 2: provide details on Acme Opportunity.
  • API call 3: provide details on champion John Doe Prospect.
  • API call 4: provide a list of calls that happened between days X–Y.
  • API call 5: provide details on any emails sent to the Acme Account, Opportunity, and Prospect.
  • And so on…

The number of API calls made is highly dependent on how Claude reasons, which depends on Claude itself and how the customer has configured Claude or their agent.

Each API call that Claude or an agent makes is counted as 1 API call to the Outreach MCP endpoint. The actual number of calls is non-deterministic, and Outreach can’t predict or provide firm guidance on it.

When searching Kaia meetings via MCP, is access limited to meetings the user is involved in?

Yes. Access is governed by the user’s profile permissions. Users can only search and retrieve meetings they are authorized to view based on their assigned permissions.