Spekit MCP Overview & Setup

1. What is the Spekit MCP Connector?

Your team already spends the day inside an AI assistant. Until now, that assistant had no view into what Spekit knows, so reps left their flow to find an approved answer, and anything drafted in a conversation stayed in that conversation.

The connector closes that gap. The same knowledge, the same permissions, and the same governance Spekit already provides are available in the AI tool your team is already using.

One Connector, Every AI Tool

You connect once, with one server URL, and the same connector works across every AI tool that supports custom connectors. You do not need a separate integration project per platform.

2. What You Can Do

The connector gives your AI tool a set of read tools and write tools. Every tool runs as you, scoped to your existing Spekit role.

🔎 Find and Use Content (Read Tools)

Get approved answers and browse your library without switching apps.

✍️ Create and Manage Content (Write Tools)

Draft and update governed content from a conversation, with no copy and paste back into Spekit.

Working Within Your Permissions

Every call runs as you. What you can see and do through the connector matches what your Spekit account can already see and do. This is enforced at the backend, not just in the connector. A viewer-only user cannot create or change content through an AI tool, even by asking.

One request can run several steps. For example, a request to build a launch content set can read your approved messaging, then draft and create each piece in turn, in order, from that one request.

3. Who Can Use It

Access depends on your AI tool, and in some tools your plan or admin role. In all cases you sign in with your existing Spekit login. The details for each tool are in the Setup section below.

• Claude: Team and Enterprise need an Owner or Primary owner to add the connector first. Pro and Max users can add it themselves.
• ChatGPT: Available on Plus, Pro, Business, Enterprise, and Education. Write tools are limited to Business, Enterprise, and Education.
• Gemini CLI: Requires Gemini CLI installed and a terminal. This is a developer-tool connection, not a settings screen.
• Salesforce Agentforce: Requires a Salesforce admin or developer.
• Gong: Requires a Gong Tech admin.

4. Setup

🔗 The Server URL

https://mcp.spekit.co/mcp

This one URL works across every AI tool that supports custom connectors. You do not need a Client ID or Client secret. You authorize with a standard OAuth sign-in using your Spekit login.

The General Setup Pattern

The steps are similar in every tool:

1. Open your AI tool's connector or integration settings. Look for "Connectors," "Custom connectors," "MCP servers," or "Integrations."
2. Add a custom connector.
3. Paste in the Spekit server URL. If the tool asks for a name, use "Spekit."
4. Sign in with your existing Spekit credentials to authorize.

A few things vary by tool. Some need a developer setting turned on first. Some are admin managed. A few use a command line or config file instead of a settings screen. None of this changes the URL, the sign-in, or your Spekit permissions.

Connect to Claude

Who can use this

• Team and Enterprise plans: an Owner or Primary owner adds the connector at the organization level first.
• Pro and Max plans: any user can add it individually.

You need an active Spekit account and a Claude plan that supports connectors (Team, Enterprise, Pro, or Max).

Team and Enterprise plans

Step 1. Owner or Primary owner adds the connector:

• Go to Organization settings, then Connectors.
• Click Add.
• Hover over Custom, then select Web.
• Enter the Spekit server URL.
• Click Add to finish.

Step 2. Members connect their own account:

• Go to Settings, then Customize, then Connectors.
• Find the Spekit connector, labeled "Custom."
• Click Connect, then sign in with your Spekit credentials.

Pro and Max plans

1. Go to Customize, then Connectors.
2. Click the plus sign, then Add custom connector.
3. Enter the Spekit server URL and click Add.
4. Select the Spekit connector and click Connect.
5. When the "Application Access Request" prompt appears, click Allow Access and sign in with your Spekit credentials.

Connect to ChatGPT

Who can use this Custom MCP connectors are available on ChatGPT Plus, Pro, Business, Enterprise, and Education plans. For Spekit, write tools are currently limited by plan:

• Business, Enterprise, and Education can use both read and write tools.
• Pro may be able to use the read tools but not the write tools.
• Plus cannot use custom MCP connectors.

You need an active Spekit account, a plan from the list above, and Developer Mode turned on.

Step 1. Turn on Developer Mode

• Open Settings.
• Go to Connectors, then Advanced settings. Depending on your account, this may appear under Security and login.
• Turn on Developer mode.

Step 2. Add Spekit as a custom connector

• In Settings, then Connectors, click Add custom connector.
• Enter a name, for example "Spekit," and the Spekit server URL, including the /mcp path.
• Create the connection and review the tools it finds.

Step 3. Authenticate. When prompted, sign in with your Spekit credentials.

Step 4. Use it in a conversation. Select the Spekit connector as a source for your conversation.

Connect to Gemini CLI 

Who can use this This is a developer-tool connection, not a point-and-click settings page. You need Gemini CLI installed and a terminal.

You need Gemini CLI installed and an active Spekit account. Gemini CLI's OAuth flow signs you in through your existing Spekit login in a browser window.

Step 1. Add the server to your Gemini CLI settings Add an entry to ~/.gemini/settings.json (global) or .gemini/settings.json in your project directory:

{
  "mcpServers": {
    "spekit": {
      "url": "https://mcp.spekit.co/mcp",
      "oauth": { "enabled": true }
    }
  }
}

Step 2. Authenticate In Gemini CLI, run:

/mcp auth spekit

This opens a browser window for you to sign in with your Spekit credentials.

Step 3. Confirm it is connected Run /mcp auth with no arguments to list servers and confirm Spekit shows as authenticated.

⚠️ A note on headless environments: The sign-in step opens a real browser window. It will not work over a remote SSH session or in a headless environment without local browser access.

Connect to Salesforce Agentforce

⚠️ Admin setup required. This is not a self-serve connection. You need a Salesforce admin or developer with access to Setup, the Tooling API (or Salesforce CLI or Workbench), and permission set assignment.

Prerequisites

• An Agentforce-enabled Salesforce org with an existing agent.
• That agent set up for multi-agent orchestration (Step 1).
• No Client ID or Client secret on Spekit's side. Salesforce creates its own credential objects when you register the server.

Step 1. Confirm your agent supports MCP tool registration Agentforce currently needs your agent's planner set to concurrent multi-agent orchestration before it can register external MCP tools. If it is not already set:

• Query your agent's GenAiPlannerDefinition record.
• Update its PlannerType field to Atlas__ConcurrentMultiAgentOrchestration, using Salesforce CLI, the Tooling API, or Workbench.

Step 2. Register the Spekit MCP server

• In Setup, search for Agentforce Registry.
• Click New.
• Enter the Spekit server URL and follow the wizard to select which Spekit tools to register.

Salesforce automatically creates a Named Credential, an External Credential, and a Permission Set for this connection.

⚠️ Don't skip Step 3. The connection will not work without it.

Step 3. Assign the permission set Assign the permission set Salesforce just created to the users, or the agent's running user, who need access.

Step 4. Add the Spekit tool to your agent

• Open your agent's configuration.
• Create or select a Topic.
• Under This Topic's Actions, click Add from Asset Library.
• Select the registered Spekit MCP tool or tools.

Testing Test inside the Agentforce builder. If a tool call does not behave as expected, inspect the mcp_request and mcp_response fields for debugging.
 

Connect to Gong

Who can use this You need a Gong Tech admin role. This is available on any Gong plan.

This setup runs in the reverse direction from the others. Here Gong itself connects out to Spekit's MCP server as an external data source, so "Ask Gong" can use Spekit content.

📌 Skip the "Client access" fields

Gong's setup screen offers an optional Client ID and Client secret section. Spekit's server does not need it. Leave it collapsed and use Personal access instead (Step 3).

Step 1. Go to MCP connections

• In Gong, go to Admin center.
• Go to Settings, then MCP connections, under Ecosystem.
• Click New connection.

Step 2. Enter connection details

• Connection name: Spekit
• MCP server URL: the Spekit server URL
• Authentication requirement: enabled

Step 3. Choose Personal access, not Shared access Gong offers Shared access (one token for all users) and Personal access (each user authorizes separately). Use Personal access. Spekit's permissions are scoped to each user's role, and Shared access would collapse every Gong user into one Spekit identity.

Step 4. Connect Click Connect and complete Spekit sign-in when prompted. Each user authorizes individually the first time.