Optimal Access · Docs

Connect Kurator to Claude and/or chatGPT

Search everything you've bookmarked and save new sources — right inside a Claude or ChatGPT conversation.

Once connected, you can ask Claude or ChatGPT to search your Kurator library and save new pages into your folders — with clean descriptions and tags — without ever leaving the chat.

The connector is built on the Model Context Protocol (MCP), the open standard that AI assistants like Claude and ChatGPT use to connect to outside tools. It signs in with secure OAuth: your AI assistant never sees your password, your data stays in your own Kurator account, and the connector only ever touches the account you sign into.

What You Can Do With It

  • Read and list your folders, prompts, create folders, generate description, list tags
  • Use your custom prompts to generate and save description
  • Read files on your google drive, apply custom prompts to generate descriptions and save them to Kurator
  • List and use posts based on tags and specific folders

Safe by design. The connector can only read your library and add to it (new folders and links). It has no ability to delete or overwrite anything in your Kurator account.

Before You Start

  • A Kurator account. You'll sign into it during setup.
  1. Claude or ChatGPT with connector support.
    • Claude: the Claude desktop app or Claude on the web (claude.ai). Connectors live in Claude's settings.
    • ChatGPT: ChatGPT on the web or desktop app. Connectors live in ChatGPT's settings.
  2. Optional — your own AI key set in Kurator (Settings → My AI Settings). If it's set, descriptions are generated server-side with your key and your saved prompt, exactly like the Kurator extension. If not, Claude or ChatGPT writes the description itself from the page. Either way works.

Connect to Claude

Works the same in the Claude desktop app and on claude.ai.

1. Open the connectors settings
In Claude, go to Settings → Connectors.

2. Add a custom connector
Click Add custom connector, give it a name (for example, Kurator), paste the URL below, and click Add.

Connector URL:
https://mcp.optimalaccess.com/mcp

3. Sign in to Kurator
Click Connect on the new connector. A Kurator sign-in page opens. Log into the account you want to use. This is the account the connector will read from and save to.

4. Approve access
Kurator shows a consent screen requesting access to your account. Review it and click Agree. You'll return to Claude, and the connector shows as Connected.

5. Confirm it works
In a new chat, ask: Confirm my Kurator account. Claude reports the account name and email it's connected to. If that's the right account, you're all set.

Approving tools: The first time Claude uses an action, it asks permission. The read-only ones (confirm account, list folders, list tags, search) are safe to set to Always allow. The actions that write to your library (create a folder, save a link) will ask each time unless you allow them too.

Connect to OpenAI

1. Open Plugins
In ChatGPT, go to Plugins.

2. Create a custom MCP server
Click Add in the top right, then choose Create custom MCP server.

3. Enter the connector details
Give it a name (for example, Kurator), paste the URL below, and click Create.

Connector URL:
https://mcp.optimalaccess.com/mcp

4. Sign in to Kurator
A Kurator sign-in page opens. Log into the account you want to use. This is the account the connector will read from and save to.

5. Approve access
Kurator shows a consent screen requesting access to your account. Review it and click Agree. You'll return to ChatGPT, and the connector is ready to use.

6. Confirm it works
In a new chat, ask: Confirm my Kurator account. ChatGPT reports the account name and email it's connected to. If that's the right account, you're all set.

Approving actions: ChatGPT may ask for confirmation before using an action, especially ones that write to your library (create a folder, save a link). Read-only actions like search and list folders are safe to approve freely.

Use Case

Use Kurator with Claude or ChatGPT for job search, prospecting, competitor research and more

Here's an example workflow.

The goal: find people on LinkedIn posting about their AI research workflows (Claude, ChatGPT, NotebookLM and similar tools), then save each post to Kurator with a tailored outreach plan as its description.

Step 1: Run a Boolean search

Ask Claude or ChatGPT to search the web with a query like this one:

site:linkedin.com/posts
(ChatGPT OR Claude OR Gemini OR NotebookLM OR Perplexity)
(research OR workflow)
(sources OR references OR bookmarks OR links OR citations OR context)

Step 2: Save the results with a custom prompt

Tell Claude or ChatGPT to save the results to a specific Kurator folder (for example, LinkedIn Outreach) and to use a custom prompt for the descriptions. If you've set your own AI key in Kurator, the description is generated with your saved prompt. If not, the assistant applies the prompt itself.

Here's the prompt I use for my own workflow. You should obviously replace information on Kurator with your own profile.

LinkedIn Outreach Prompt

You are an expert technical product strategist and B2B outreach advisor for Kurator.

Analyze the content, which details a person's AI research workflow (e.g., using Claude, ChatGPT, NotebookLM, local context folders, or custom instructions). Based on this content, generate a personalized outreach strategy. Structure your response as follows:

0. ABOUT KURATOR (reference — do not output this section)
Kurator captures web pages and AI research sessions as individual saved posts, organized into collection folders. It works through a Chromium browser extension and through a connector inside Claude and ChatGPT, so users can search their library and save new sources without leaving the chat. When saving, the user specifies:
* the search terms or content to capture
* the collection folder where the link is stored
* the custom prompt used to generate the saved description (like the one you are reading now)

1. WORKFLOW INTERSECTION
* Briefly identify how their current AI workflow (Claude, ChatGPT, or other tools) aligns with Kurator.
* Highlight the missing gap: how Kurator bridges their research sessions (web and AI chat) back to structured, reusable, and published/curated knowledge.

2. ANGLE & VALUE PROPOSITION
* Framing: position Kurator not as a replacement for their AI assistant, but as the bookmarking, annotation, and prompt-execution layer for it.
* Key pitch point: show how they can research directly in Claude or ChatGPT, save the output and its sources to Kurator from inside the chat, and run a custom prompt against that content at the moment of saving.

3. OUTREACH MESSAGING OPTIONS
Option A: The "Compliment & Extend" Approach (Soft Connection)
* A friendly, value-first message acknowledging their post or setup.
* Focuses on how Kurator complements their existing setup (local Markdown notes, NotebookLM sources, custom instructions) by handling web capture, metadata, and custom prompt execution on saved notes.

Option B: The "Workflow Upgrade" Approach (Direct Pitch)
* A concise pitch focused on saving AI outputs and web sources directly into curated collections, with automated prompt execution on save.

4. SUGGESTED NEXT STEP
* A low-friction call to action (e.g., asking for feedback on a quick 30-second demo, or sharing a template).