Back to MCP setup

ChatGPT Setup

Connect ContextDrop to ChatGPT

Connect ChatGPT to ContextDrop through MCP and access persistent project knowledge shared across your AI clients.

Custom MCP server availability and supported capabilities depend on your ChatGPT plan and account settings. The interface may differ from the screenshots in this guide.

Before you start

What you need

  • A ContextDrop account
  • Access to custom MCP server connections in ChatGPT
  • A project with memories in ContextDrop

Step 1

Open ChatGPT Plugins

Open ChatGPT's Plugins page. Click the Add button and select Add custom MCP server.

If the option isn't available, check whether your account supports custom MCP connections.

Open ChatGPT Plugins ↗

Step 2

Configure the connection

Enter the following details in the Add custom MCP server form.

Name

ContextDrop

Description

Persistent project memory shared across AI clients. Retrieve, search, save, and update project knowledge.

Connection

Server URL

Server URL

https://mcp.usecontextdrop.com/mcp

Authentication

OAuth

Step 3

Configure OAuth

Select OAuth as the authentication method and expand Advanced OAuth settings. ChatGPT should discover ContextDrop's OAuth endpoints automatically.

Select User-Defined OAuth Client

In our tested setup, Dynamic Client Registration (DCR) was selected by default, but ChatGPT reported that the server did not advertise CIMD support. We selected User-Defined OAuth Client instead.

Leave the discovered OAuth endpoints unchanged. Review the available scopes and settings, acknowledge the custom-server warning if you trust the server, and click Create as a plugin.

Step 4

Authenticate with ContextDrop

After creating the plugin, ChatGPT displays a connection prompt. Click Connect ContextDrop to open the OAuth flow.

  1. 1. Click Connect ContextDrop.
  2. 2. Sign in to your ContextDrop account.
  3. 3. Authorize ChatGPT when prompted.
  4. 4. Confirm the connection completes successfully.

Step 5

Select a project and configure instructions

In ContextDrop, open Settings → MCP & Integrations and select the project you want to use as your active MCP workspace.

To encourage autonomous memory retrieval and capture, you can also add ContextDrop instructions in ChatGPT's Custom Instructions. This is optional for basic MCP connectivity and retrieval.

Recommended instructions

Use ContextDrop as the persistent memory layer for project-specific work.

Retrieval:
- At the start of project-specific tasks, search ContextDrop for relevant memories.
- Treat retrieved memories as supporting context and check for conflicts.
- If nothing relevant is found, continue normally.

Memory maintenance:
- Identify durable knowledge such as architecture, decisions, conventions, requirements, and technical discoveries.
- Search existing memories before saving to avoid duplicates.
- Save new durable knowledge and update existing memories when established knowledge changes.
- Ignore temporary debugging details, speculation, secrets, and irrelevant information.
- Never claim a save or update succeeded unless the tool confirms it.

Conversational continuity:
- Keep memory operations unobtrusive.
- Continue the user's original task immediately after a successful operation.
- Do not announce routine searches or successful saves.
- Explain memory operations only when asked or when a meaningful conflict, failure, or clarification requires attention.

Security:
- Use only the authenticated user's authorized ContextDrop workspace.
- Never attempt to access another user's or workspace's memories.

ChatGPT Custom Instructions have a 1,500-character limit for this field. The instructions above are designed to fit, but verify the count if you edit them.

Step 6

Verify the connection

Start a new chat and make sure ContextDrop is enabled for that conversation. Ask ChatGPT to retrieve an existing project memory that was previously saved by another connected AI client.

Example verification prompt

Search ContextDrop for the API validation and authorization convention established in another AI client.

Retrieve the saved memory and summarize the standards I should follow when implementing an API endpoint. Don't save anything new yet.

Successful verification

  • ChatGPT invokes ContextDrop's memory tools.
  • It retrieves the memory saved by the other AI client.
  • Its answer reflects the retrieved project knowledge.

Secure OAuth authentication

ContextDrop uses OAuth to authorize ChatGPT. Sign in through ContextDrop and review the authorization prompt before granting access. You should not share your ContextDrop password or tokens in a chat.

← Back to MCP setup