Connect your AI tools in a few minutes.
Pick the way that fits the tool you use. Each one has step-by-step directions. No technical setup, and you can undo any connection at any time.
Pick the way that fits the tool you use. Each one has step-by-step directions. No technical setup, and you can undo any connection at any time.
This is the easiest way. The tool asks you to sign in to AI Context Keeper, you click Approve, and you are done.
Your connector address
https://www.aicontextkeeper.com/mcpAI Context Keeper.https://www.aicontextkeeper.com/mcp.How you know it worked: start a new chat and type "List my AI Context Keeper workspaces." Claude lists them.
If it didn't: make sure the connector is turned on for the chat. Click the plus or tools button in the chat box, find AI Context Keeper, and switch it on. If Claude says the connection failed, remove the connector in Settings › Connectors and add it again.
Connectors you add on the web show up in the Claude desktop and mobile apps on the same account. There is nothing extra to install.
How you know it worked: type "List my AI Context Keeper workspaces."
If it didn't: quit the app fully and open it again. If the connector is still missing, sign out and back in.
Claude Code is Anthropic's tool for the terminal. You add the connector with one command, then sign in.
claude mcp add --transport http aicontextkeeper https://www.aicontextkeeper.com/mcpclaude and pressing Enter./mcp and press Enter. Pick aicontextkeeper from the list and choose to sign in.How you know it worked: in Claude Code, type "List my AI Context Keeper workspaces."
If it didn't: run claude mcp list to check the server is there. If it is missing, run the add command again. If sign-in keeps failing, remove it with claude mcp remove aicontextkeeper and start over.
Prefer a token instead of signing in? See "Install on your computer" below. Claude Code works either way.
ChatGPT connects through a setting called Developer Mode. It is a one-time switch; you do not need to be a developer.
AI Context Keeper.https://www.aicontextkeeper.com/mcp.How you know it worked: type "List my AI Context Keeper workspaces."
If it didn't: Developer mode may not be available on every ChatGPT plan. Check that the switch in step 4 stayed on. If ChatGPT says the connection failed, delete the connector and create it again.
Many AI tools let you add a connector by pasting an address. The words differ a little from tool to tool, but the steps are the same.
AI Context Keeper.https://www.aicontextkeeper.com/mcp as the address or URL.How you know it worked: ask the tool to list your AI Context Keeper workspaces.
If it didn't: some tools do not support signing in this way. Use "Install on your computer" with a token, or "Share a link".
When a tool connects, AI Context Keeper shows you a page before anything happens. It names the tool, shows which account you are signed in with, and lists two permissions:
Approve grants both permissions. What the tool is allowed to do with them, and whether it asks you first, is set inside the tool. In Claude, that is Settings › Connectors › Configure.
Click Approve to connect, or Deny to stop. To remove a tool later, go to Account › API, find it under Connected apps, and click Revoke. The tool loses access right away.
Some tools run on your own machine and need a small helper program to reach your files. Setup is two steps: make a token, then paste one block of text into the tool. The helper installs itself the first time the tool starts.
You need Node.js version 20 or newer on your computer. Most of these tools already require it. If you are not sure, the tool's own install guide will say.
A token is a long password that lets one tool reach your account. You make one per tool.
Cursor.ack_. Click Copy and paste it somewhere safe, such as a password manager. You will not see it again.Keep the token private. Anyone who has it can read your files. If you think it has leaked, come back to this tab and click Revoke.
In every block below, replace your-token with the token you just copied. Keep the quotation marks.
claude_desktop_config.json opens in a text editor, or the folder that holds it opens.{
"mcpServers": {
"aicontextkeeper": {
"command": "npx",
"args": ["-y", "@aicontextkeeper/mcp-server"],
"env": { "ACK_API_TOKEN": "your-token" }
}
}
}"aicontextkeeper" part inside the existing "mcpServers" section, after a comma.your-token with your token and save the file.Where the file lives, if you need to find it yourself:
~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.jsonHow you know it worked: start a new chat, click the tools button, and AI Context Keeper is listed. Type "List my AI Context Keeper workspaces."
If it didn't: the most common cause is a missing comma or bracket in the file. Copy the block again and check that every { has a matching }. Then restart the app.
your-token, and press Enter:claude mcp add aicontextkeeper --scope user --env ACK_API_TOKEN=your-token -- npx -y @aicontextkeeper/mcp-serverHow you know it worked: run claude mcp list. AI Context Keeper is listed. In Claude Code, type "List my AI Context Keeper workspaces."
If it didn't: run claude mcp remove aicontextkeeper and try the command again, checking the token has no spaces.
mcp.json opens.your-token:{
"mcpServers": {
"aicontextkeeper": {
"command": "npx",
"args": ["-y", "@aicontextkeeper/mcp-server"],
"env": { "ACK_API_TOKEN": "your-token" }
}
}
}How you know it worked: open the chat panel and type "List my AI Context Keeper workspaces."
If it didn't: click the refresh icon next to the server in Settings. If the dot stays red, check the file for a missing comma or bracket.
~/.codeium/windsurf/mcp_config.json opens. Paste the same block as Cursor above, replacing your-token.How you know it worked: in the chat panel, type "List my AI Context Keeper workspaces."
If it didn't: check the file for a missing comma or bracket, then refresh.
your-token, and press Enter:codex mcp add aicontextkeeper --env ACK_API_TOKEN=your-token -- npx -y @aicontextkeeper/mcp-serverHow you know it worked: start Codex and type "List my AI Context Keeper workspaces."
~/.gemini/settings.json in a text editor. Create it if it does not exist.your-token. If the file already has settings, add the "mcpServers" section inside the outer braces, after a comma.How you know it worked: type /mcp to see the server listed, then "List my AI Context Keeper workspaces."
opencode.json in your project, or ~/.config/opencode/opencode.json).your-token:"mcp": {
"aicontextkeeper": {
"type": "local",
"command": ["npx", "-y", "@aicontextkeeper/mcp-server"],
"environment": { "ACK_API_TOKEN": "your-token" }
}
}How you know it worked: type "List my AI Context Keeper workspaces."
The npx -y part of every block fetches the current helper each time the tool starts. There is nothing to update by hand.
Any tool that can read a web page can read a link from AI Context Keeper. There is nothing to install and nothing to sign in to. Use this for Perplexity, Grok, Gemini in the browser, and any tool not listed above.
The tool opens the link and reads your file as plain text.
By default a link works until you delete the file or move it to Trash.
After that time the old link stops working. The next time you copy a link, you get a new one.
A link needs no login, so anyone who has it can read that file. Share it the way you would share any link.
The Chrome extension goes the other way. It saves good answers from your AI tools into AI Context Keeper, so your files stay current.
How you know it worked: the file appears in your workspace. Any connected tool can read it right away.
Models that run on your own computer connect the same way as the tools above.
your-token with a token from Step A.How you know it worked: type "List my AI Context Keeper workspaces."
https://www.aicontextkeeper.com/mcp.AI Context Keeper and click Save.How you know it worked: type "List my AI Context Keeper workspaces."
Ollama runs the model. To give it your files, use it through a front end that supports connectors, such as Open WebUI (steps above) or a coding tool from "Install on your computer" that lets you pick an Ollama model.
Once connected, a tool can:
Things to try:
Connect your first tool in a few minutes. Change a file once and every tool sees it.