> For the complete documentation index, see [llms.txt](https://docs.spara.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.spara.com/guides/integration-guides/connect-spara-to-claude-or-chatgpt.md). # Connect Spara to Claude or ChatGPT Connect Claude or ChatGPT to your Spara workspace to inspect, analyze, and safely draft changes to your agents. Connect Spara to Claude or ChatGPT to explore your Agents and channels, analyze performance, and prepare configuration changes for review. You can also ask product questions using Spara's public documentation. ## Before you connect You need: * A Spara account and membership in the workspace you want to use * A supported client: the Claude app or Cowork, or ChatGPT on the web or in the desktop app * An Editor or Manager role in Spara if you want to create Agents or channels, or draft configuration changes. Members and Integrators can use the read and analytics tools. See [User Management](https://docs.spara.com/platform/settings/user-management) for role details. ChatGPT web and desktop use different setup paths. On the web, connect directly to Spara's MCP server. In the desktop app, install Spara from its plugin marketplace. Availability depends on your ChatGPT plan, workspace permissions, and selected interface. For managed AI workspaces, an administrator may need to enable Spara before members can connect: * **ChatGPT managed workspaces**: An administrator may need to allow developer mode and custom MCP connections for web setup, or import the Spara marketplace and make the desktop plugin available to your role. * **Claude Team and Enterprise**: A Claude Owner or Primary Owner must first add Spara for the Claude organization. ## Install and connect Spara Spara's public plugin repository is [spara-ai/spara-plugins](https://github.com/spara-ai/spara-plugins). Use `spara-ai/spara-plugins` for marketplace installation, not `spara-ai/spara-app`. {% tabs %} {% tab title="ChatGPT web" %} Connect directly to Spara's MCP server using ChatGPT developer mode. This gives ChatGPT access to Spara's tools without installing the desktop plugin bundle. {% stepper %} {% step %} ### Enable developer mode Open [ChatGPT](https://chatgpt.com), then **Settings > Security and login**, and turn on **Developer mode**. If this option is unavailable, ask your ChatGPT workspace administrator about access to custom MCP connections. {% endstep %} {% step %} ### Create the Spara connection Open [ChatGPT Plugins](https://chatgpt.com/plugins) and select the **+** button to add an MCP server. Name it **Spara**, add a short description, and enter this server URL: ``` https://app.spara.co/mcp/customer/stream ``` Use **OAuth** authentication and complete the connection details. Select **Create** and follow the prompts to sign in to Spara, select your Spara workspace, and authorize access. {% endstep %} {% step %} ### Install and test the connection Open [Personal plugins](https://chatgpt.com/plugins?view=personal), find the Spara connection you created, and select **+** to install it. Complete Spara sign-in if prompted. Return to ChatGPT, switch to **Work**, and start a new chat. Type `@`, select **Spara**, and ask: "What agents and channels do we have set up in Spara?" Check that the results match the workspace you selected. {% endstep %} {% endstepper %} These steps follow OpenAI's [MCP plugin quickstart](https://developers.openai.com/plugins/quickstart). The GitHub marketplace plugin described in the desktop tab is **Desktop only**, even though it connects to a remote server. {% endtab %} {% tab title="ChatGPT desktop" %} Install the Spara marketplace plugin in the ChatGPT desktop app. It includes Spara's MCP connection and reusable instructions for working with Spara. {% stepper %} {% step %} ### Add the Spara marketplace For a managed ChatGPT workspace, ask an administrator to open **Admin > Plugins > Add > Import marketplace** and enter this repository URL as **Source**: ``` https://github.com/spara-ai/spara-plugins ``` Leave **Path** and **Branch, tag, or commit** empty to use the repository root and default branch. Select **Import marketplace**, authorize GitHub access if prompted, and make Spara **Available** or **Installed** for the intended roles. Importing the marketplace does not connect each member's Spara account. If you manage plugins locally with Codex CLI instead, add the same marketplace with: ```sh codex plugin marketplace add spara-ai/spara-plugins ``` Then reopen the desktop app's **Plugins** tab to find the Spara marketplace. See OpenAI's [local marketplace setup](https://developers.openai.com/plugins/build/plugins#add-a-marketplace-from-the-cli) for that alternative. {% endstep %} {% step %} ### Install and authorize Spara In the ChatGPT desktop app, open **Plugins**, find **Spara** in the workspace or configured marketplace, and select **+** to install it. Connect the Spara MCP server when prompted. Sign in to Spara, select the workspace you want to use, and approve access. The plugin supplies the server URL; you do not need to enter it manually. {% endstep %} {% step %} ### Start a new conversation Start a new Work chat and ask Spara: "What agents and channels do we have set up in Spara?" Complete the connection if authentication is requested on first use, and check that the results match your selected workspace. {% endstep %} {% endstepper %} OpenAI documents this flow in [Plugins](https://learn.chatgpt.com/docs/plugins) and [Plugin management](https://learn.chatgpt.com/docs/enterprise/plugin-management). Imported plugins with an MCP server are **Desktop only**. Importing Spara into your workspace does not make that plugin usable on the web; use the separate web setup above. {% endtab %} {% tab title="Claude" %} {% stepper %} {% step %} ### Add the Spara marketplace Open **Customize > Plugins > Add > Add Marketplace**. Copy and paste this URL into the dialog: ``` https://github.com/spara-ai/spara-plugins ``` {% endstep %} {% step %} ### Install Spara Find the Spara plugin card and select the **+** button in its top-right corner. {% endstep %} {% step %} ### Connect your Spara account Select **Settings > Connectors > Connect** next to the Spara connector. Sign in to Spara and select the workspace you want Claude to use. {% endstep %} {% endstepper %} {% endtab %} {% endtabs %} ## What you can do All access is limited to the Spara workspace selected during OAuth. Spara checks your current workspace membership and permissions for every tool call. ### Read and analyze Any member of the selected workspace can ask Claude or ChatGPT to: * List Agents and their Chat, Email, Phone, SMS, Product Demo, and Workflow channels * Read channel configuration and active AI Instructions * List Spara Data Model fields and available fields from a connected CRM * Find individual Leads and read their Spara Data Model values * Find conversations by Lead, topic, channel, or recency, and read transcripts from Chat, Email, SMS, Phone, and Product Demo conversations * List voices and Product Demo features used when configuring channels * Analyze trends, comparisons, and performance questions over the workspace's Spara data * Search and read Spara's public documentation Analytics returns a written answer. It does not expose raw SQL or raw result rows. ### Create and draft changes Editors and Managers can ask Claude or ChatGPT to: * Create an Agent container * Create an unpublished channel * Update a channel's configuration * Edit an Agent's AI Instructions These configuration tools are hidden and unavailable when your current Spara role does not include full Agent access. ChatGPT workspace controls can also disable actions or require you to approve them. Any workspace member can submit product feedback through the connector. Feedback is recorded for Spara's product team and does not change the workspace. ## Changes never publish automatically The connector has no tool that can publish a channel or configuration change. Changes to channel configuration and AI Instructions are saved as **unpublished drafts**. New channels also start unpublished. Your AI assistant returns a link to the relevant editor in Spara, where a person must review and publish the draft. Until that happens, the published configuration remains unchanged. Creating an Agent container or submitting product feedback does not publish a live channel change. ## Data handling and telemetry Your AI assistant can request only the data exposed by the capabilities above. Depending on the request, this can include Agent and channel configuration, Data Model and CRM field definitions, individual Lead details, conversation transcripts, and analytics derived from Lead and conversation data. Analytics may process underlying Lead and conversation data, but it returns only a written analysis. The analytics tool receives both the question and a short natural-language description of why the question is being asked. Spara uses that context for observability and product improvement; it does not change the analysis. Spara records connector tool calls for security, reliability, evaluation, and product usage monitoring: * Spara stores a redacted and size-limited copy of tool inputs and a redacted result preview of up to 20,000 characters. * Product telemetry includes the selected workspace, tool name, outcome, duration, OAuth client identifier, and error type when applicable. * Diagnostic monitoring includes the Spara server and tool involved in an error. See Spara's [Privacy Statement](https://www.spara.com/privacy) and [Developer: Security & Compliance](https://docs.spara.com/developers/installation/security-and-compliance) for Spara's broader data protection, storage, encryption, and AI subprocessor commitments. For ChatGPT-specific controls and data handling, see OpenAI's [Apps in ChatGPT](https://help.openai.com/en/articles/11487775). ## Disconnect, reconnect, or reset access {% tabs %} {% tab title="ChatGPT" %} 1. Open **Settings**, then select **Plugins** or **Apps**, depending on the option shown for your account. 2. Open Spara. Under **Connected accounts** or **Connection**, open the more options menu and select **Disconnect**. To reconnect or reset an expired token, return to Spara's plugin listing and select **Reconnect** or **Connect**. Sign in again and select the intended Spara workspace. See OpenAI's [Connecting and managing app accounts in ChatGPT](https://help.openai.com/en/articles/20001494-connecting-and-managing-app-accounts-in-chatgpt) for account-management details. {% endtab %} {% tab title="Claude" %} 1. Open **Customize > Plugins**. 2. Find **Spara**, open its more options menu, and select **Remove**. To reconnect or reset an expired token, install Spara again and complete the connection flow. {% endtab %} {% endtabs %} Disconnecting prevents future Spara access through that connection. ## Troubleshooting ### Spara does not appear * **ChatGPT web**: Create and install the direct MCP connection from the web tab above. A GitHub-imported plugin marked **Desktop only** cannot run on the web. If developer mode is missing, ask your workspace administrator about custom MCP access. * **ChatGPT desktop**: Confirm that the Spara marketplace was added from `https://github.com/spara-ai/spara-plugins`. In a managed workspace, ask an administrator to check the import result and installation policy for your role. Start a new chat after installation. * **Claude**: Confirm that the Spara marketplace and plugin were added in **Customize > Plugins**. On Team and Enterprise plans, ask a Claude Owner to confirm that Spara is enabled for the organization. ### Authentication succeeds, but tool calls are denied Disconnect and reconnect, then select a Spara workspace where you are still a member. Spara checks live membership on every call, so removed or revoked access takes effect immediately. ### Configuration tools are missing Your Spara role does not have full Agent access. Ask a Manager to assign you the Editor or Manager role in [**Settings > Team**](https://app.spara.co/settings/team), then reconnect so the assistant refreshes the visible tools. ### The assistant is using the wrong Spara workspace Disconnect and reconnect, then select the correct workspace during OAuth. One connection is scoped to one workspace at a time. ### You still cannot connect Email with the client and mode you are using, the approximate time of the failure, and the error message. Do not send access tokens or passwords. ## Current limitations * The connector cannot run raw SQL or return raw analytics rows. * The connector cannot publish or delete Agents or channels, edit Lead records, or manage CRM field mappings. * One OAuth connection is scoped to one Spara workspace. Reconnect to select another workspace. * The GitHub-imported Spara plugin is desktop-only. ChatGPT web requires a separately created MCP connection. Workspace policies still apply to both setup paths. * The Spara plugin is not yet listed in Anthropic's official plugin marketplace. Claude installation depends on the public Spara plugin repository. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.spara.com/guides/integration-guides/connect-spara-to-claude-or-chatgpt.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.