> For the complete documentation index, see [llms.txt](https://docs.cogram.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cogram.com/integrations/connecting-ai-apps.md).

# Connecting AI apps

Connect ChatGPT, Claude, or another AI app to Cogram to ask about your projects and make changes in them: meeting minutes, emails, RFIs, submittals, drawings, and field reports.

Connect your AI app to Cogram once, and from then on you can ask it about your projects directly — "what did we decide about the curtain wall in last week's OAC meeting?", "list open RFIs on the Riverside project", "summarize this month's field reports". You can also ask it to make changes, such as "add a task to the Riverside board to chase the door hardware submittal".

Three facts define what the AI can do:

* **It acts as you.** The AI sees what you can see in Cogram and can change what you can change, and nothing more.
* **Your organization sets the limits.** Your administrator's Tool access settings for Cogram Agent apply to connected AI apps too.
* **Some changes wait for your approval.** For those, nothing changes until you approve the request in Cogram.

This page covers ChatGPT and Claude step by step, but any AI app that supports MCP connectors works the same way — see [Other AI apps](#other-ai-apps).

## Before you start

You need your own Cogram account, in an organization where an administrator has turned on MCP access. If the connection is refused, ask your administrator — the setup is described in [MCP Server](https://docs.cogram.com/integrations/mcp-server.md).

## Connect ChatGPT

In a ChatGPT workspace, Cogram is installed as a plugin. A workspace admin must publish the **Cogram MCP** plugin to the workspace first — send them to [MCP Server](https://docs.cogram.com/integrations/mcp-server.md). Then each user installs it once:

1. In the ChatGPT sidebar, select **Plugins**.
2. Select your workspace's tab, next to **Public** — it carries your company's name.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-524391632fae5b811cf209615de5c991470717fe%2Fchatgpt-plugins-cogram.png?alt=media" alt="ChatGPT Plugins page with the workspace tab selected, showing the Cogram MCP plugin" width="700"><figcaption><p>Plugins → your workspace's tab. Cogram MCP is listed there once your admin has published it</p></figcaption></figure>

3. Select **Cogram MCP**, then select **Install plugin**.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-f4a734ad52f09d6d9a05e0372e5d7fd966bc3cdb%2Fchatgpt-cogram-mcp-install.png?alt=media" alt="Cogram MCP plugin page in ChatGPT with an Install plugin button" width="700"><figcaption><p>Select <strong>Install plugin</strong></p></figcaption></figure>

4. In the **Connect Cogram MCP** dialog, select **Continue to Cogram MCP**.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-4201d421041392197099e27b7eb8229e47d0f50e%2Fchatgpt-connect-dialog.png?alt=media" alt="ChatGPT Connect Cogram MCP dialog with a Continue to Cogram MCP button" width="700"><figcaption><p>Select <strong>Continue to Cogram MCP</strong></p></figcaption></figure>

5. A Cogram tab opens. Sign in as you normally do, including through your identity provider if your organization uses SSO, and on the approval screen — **Allow ChatGPT to access Cogram?** — select **Allow**. If you have signed in and approved before, this step completes on its own.

**Expected result:** ChatGPT shows **Cogram MCP is installed**, and the plugin page now offers **Try in chat**.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-466bb7e7f0b2ed41eda431a3460d4bdc4e36eef9%2Fchatgpt-cogram-installed.png?alt=media" alt="Cogram MCP plugin page in ChatGPT after installation, with a Try in chat button" width="700"><figcaption><p>Installed: the plugin page now offers <strong>Try in chat</strong></p></figcaption></figure>

To use it, select **Try in chat**, or in any chat select **Plugins** under the message box and pick **Cogram MCP**.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-d852e8d7a9b5ceaac59e9d589c65bae5315ba543%2Fchatgpt-plugins-picker.png?alt=media" alt="ChatGPT chat composer with the Plugins picker open, listing Cogram MCP" width="700"><figcaption><p>In a chat: <strong>Plugins</strong> under the message box → <strong>Cogram MCP</strong></p></figcaption></figure>

## Connect Claude (claude.ai)

In a Claude workspace, Cogram is a custom connector. An owner must publish the **Cogram MCP** connector to the organization first — send them to [MCP Server](https://docs.cogram.com/integrations/mcp-server.md). Then each user connects it once:

1. In the claude.ai sidebar, select **Customize**, then **Connectors**.
2. Find **Cogram MCP** in the list and select **Connect**.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-17cd13c4054e923c78e7f051de955c0e23dfd324%2Fclaude-connectors-list.png?alt=media" alt="Claude Customize → Connectors list showing the Cogram MCP connector with a Connect button" width="700"><figcaption><p>Customize → Connectors: select <strong>Connect</strong> on Cogram MCP</p></figcaption></figure>

3. A Cogram tab opens. Sign in as you normally do, including through your identity provider if your organization uses SSO, and on the approval screen — **Allow Claude to access Cogram?** — select **Allow**.

**Expected result:** you land back on the connector in claude.ai, which now shows Cogram's tools and a **Disconnect** button. On this page you can also set per-tool permissions — by default Claude asks for your approval before each tool call.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-f9b542e6457c7697d51bdcf8c4677b939b20650b%2Fclaude-connected.png?alt=media" alt="Cogram MCP connector page in claude.ai after connecting, showing tool permissions and a Disconnect button" width="700"><figcaption><p>Connected: the connector page lists Cogram's tools</p></figcaption></figure>

To use it in a conversation, open the plus menu in the chat box, select **Connectors**, and check that **Cogram MCP** is turned on — it is on by default after connecting.

<figure><img src="https://1170756420-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxpPm3rHOzFMoBqV20UQl%2Fuploads%2Fgit-blob-cc9c73394e2fd808f176d6fb57cfd3d003864892%2Fclaude-chat-connectors.png?alt=media" alt="Claude chat composer plus menu with the Connectors submenu open, showing the Cogram MCP toggle turned on" width="700"><figcaption><p>In a chat: plus menu → <strong>Connectors</strong> → <strong>Cogram MCP</strong></p></figcaption></figure>

**If you use a personal Claude account** (Free, Pro, or Max), add Cogram yourself:

1. Go to **Customize → Connectors** and select **Add custom connector**.
2. Name it **Cogram** and enter the server URL `https://mcp.cogram.com/mcp`. Leave the advanced OAuth fields empty.
3. Select **Add**, then **Connect**, and complete the same sign-in and approval as above.

Claude caveats: custom connectors work on the web, in the desktop app, and in Cowork. The Free plan allows one custom connector.

## Other AI apps

Cogram's MCP server is a standard MCP server, so any AI app that can add a custom MCP connector can connect — for example Grok, Microsoft Copilot Studio, or an agent your own team builds. The menu names differ per app, but the shape is always the same:

1. In the app, add a custom connector (sometimes called an MCP server, app, or tool) with the server URL `https://mcp.cogram.com/mcp`.
2. Sign in to Cogram in the browser window that opens, and select **Allow**.

For agents your team builds itself, see [MCP Server](https://docs.cogram.com/integrations/mcp-server.md).

## Check that it works

Ask the AI application: "list my active Cogram projects". It should call Cogram tools such as `search_projects` and answer with your projects.

## What the AI can do

**Read.** The AI can search and read projects, meetings, emails, documents, RFIs, submittals, drawings, field reports, observations, transmittals, boards, agents, and the directory. What it gets back is limited by your own project membership and [asset visibility](https://docs.cogram.com/projects/asset-visibility.md).

**Make changes.** The AI can make changes as you, for example create a project, add a board item, or create an RFI. Three things limit a change:

* **Your own permissions.** If you cannot do something in Cogram, the AI cannot do it for you.
* **Your organization's settings.** An administrator can limit or turn off areas such as tasks, documents, or projects for Cogram Agent. The same limits apply to connected AI apps.
* **Your organization's** [**licensing**](https://docs.cogram.com/organization-administration/licensing.md)**.** Tools for a module your organization is not licensed for are not offered. To make a change in a licensed module, you also need a seat for it.

**Approve.** Some changes, such as an email draft or a new project template, are not made at once. The AI tells you that the request waits for approval. Open the notification bell in Cogram, review the request, and approve or reject it. Nothing changes until you approve.

Your AI app may also ask you to confirm before it makes a change. Read what it is about to do before you allow it, and be careful with an "always allow" choice for changes.

The list of tools is in [MCP Server](https://docs.cogram.com/integrations/mcp-server.md#tools).

Anything the AI writes from Cogram data is an AI draft. Check names, dates, decisions, action items, technical values, and any contractual wording before you rely on it or let the AI save it to Cogram.

## Troubleshooting

#### MCP access is off

The AI reports "MCP access is not enabled for your organization", or the approval screen offers no **Allow** button.

* **Likely cause:** MCP access is off for your organization.
* **Fix:** Ask an administrator to turn it on — see [MCP Server](https://docs.cogram.com/integrations/mcp-server.md).

#### Cogram MCP is missing in ChatGPT

**Cogram MCP** does not appear under **Plugins** on your workspace's tab in ChatGPT.

* **Likely cause:** A workspace admin has not published the Cogram MCP plugin to your workspace yet — users cannot add it themselves.
* **Fix:** Ask a workspace admin to publish it — see [MCP Server](https://docs.cogram.com/integrations/mcp-server.md).

#### Old tools after a release

After a Cogram release, the AI still offers the old tools, even in a new chat.

* **Likely cause:** Your AI app keeps the list of tools it read when it connected. A new chat does not read the list again.
* **Fix:** In Claude, open the Cogram connector under **Customize → Connectors** and select **Refresh tools list** in its **⋯** menu. In ChatGPT, a workspace admin refreshes the list for everyone, and reinstalling the plugin does not change it. Ask your admin. In other AI apps, disconnect and connect again. You only need this for a new tool. A new operation, which the AI finds with `list_operations`, is available at once.

#### Missing projects or meetings

The AI sees fewer projects or meetings than you expect.

* **Likely cause:** The AI has exactly your own access. You are not on those projects, or asset visibility hides those records from you.
* **Fix:** Ask to be added to the projects. Nothing about the connection itself can widen what the AI sees.

#### A change is refused

The AI reports "This capability is read-only for your organization", or cannot make a change you expected.

* **Likely cause:** Your administrator's Tool access settings limit that area, or you do not have the permission or the seat for the change.
* **Fix:** Make the change in Cogram, or ask your administrator. Do not ask the AI to try another way.

#### A change stays pending

The AI says a change is pending, and nothing changed in Cogram.

* **Likely cause:** The change waits for your approval.
* **Fix:** Open the notification bell in Cogram and approve the request.

#### Asked to sign in again

A working connection suddenly asks you to sign in again.

* **Likely cause:** The stored sign-in expired or was revoked.
* **Fix:** Run the connection again — sign in and approve.

#### Disconnect the AI

You want to disconnect the AI from Cogram.

* **Fix:** Remove the Cogram connector or app in the AI application. That ends its access immediately.

## Next steps

* [Agent](https://docs.cogram.com/agent/agent.md) — Cogram's own built-in agent, which needs no setup.
* [MCP Server](https://docs.cogram.com/integrations/mcp-server.md) — administrator setup, the tools, and rate limits.

Questions? Email <support@cogram.com> or use the in-app Help menu.


---

# 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.cogram.com/integrations/connecting-ai-apps.md?ask=<question>&goal=<user_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.
