Tulina, inside the AI you already use
Copy the link below and paste it into Claude, ChatGPT or Mistral. Sign in with your Tulina account. From then on, your AI agent knows your company, follows your processes and can work in your tools, without leaving the conversation.
Last updated 8 September 2026
https://mcp.tulina.ai/mcp
One link
Nothing to install
You paste one address into your AI agent, once. There is no software to download and no setup on your side.
One sign-in
Your Tulina account
The agent asks you to sign in to Tulina, the same way you sign in to the app. What you can see and do in Tulina is what the agent can see and do.
One rule
Only what you connected
The agent reaches the tools your workspace has turned on, and nothing else. Turn a tool off in Tulina and the agent loses it too.
01
What your agent can do once connected
Everything below goes through Tulina, with the tools your workspace has turned on. A tool you have not connected is a tool the agent does not have.
Work in your tools
Look up a customer in the CRM, send an email, check a company in the registry, enrich a contact. The agent does it through Tulina, with the tools you have connected.
Follow your processes
The processes written for your company are handed to the agent. It looks for one before improvising, and runs it the way it was written.
Know your company
The agent reads your company brain and can add to it. What it learns while working is kept, so the next conversation starts further along.
Keep lists in tables
A list of leads built in a conversation lands in a Tulina table. It is there for the next conversation, and for anyone opening the app.
Leave a record
Every piece of work the agent does is logged, step by step. You can read what it did on your behalf, not only what it said.
Show, not tell
In Claude, some answers arrive as a proper table or page inside the conversation, rather than a wall of text.
02
Add it to your agent
Same link everywhere. The agent opens a Tulina sign-in window, you sign in, and that is the whole setup. Nothing else to copy.
Claude
claude.ai or the desktop app
- 1Open Settings, then Connectors.
- 2Click Add custom connector and paste the link.
- 3Click Add, then Connect, and sign in to Tulina.
Claude is the agent that shows tables and pages inside the conversation.
ChatGPT
chatgpt.com
- 1Open Settings, then Connectors, and create a connector.
- 2Paste the link and leave the other options as they are.
- 3Sign in to Tulina when ChatGPT asks.
Adding your own connector depends on your ChatGPT plan. If the option is not there, that is the plan, not Tulina.
Mistral
Le Chat
- 1Open Connectors and add a custom one.
- 2Paste the link.
- 3Sign in to Tulina in the window that opens.
Another agent may work with the same link. If it refuses the sign-in, write to hello@tulina.ai with its name and we will look.
For developers: Claude Code takes the link from the terminal, then /mcp inside a session to sign in.
claude mcp add --transport http --scope user tulina https://mcp.tulina.ai/mcp03
Agents that work on their own
Claude in Slack, or an agent that runs every morning, has nobody at the screen to sign in. Those use an access key instead, and Tulina treats the key as the person who created it.
Create an access key
In the Tulina app, open Personal settings and create an access key. It is shown once, so copy it there and then.
Give it to the agent
Paste the key where that agent keeps its credentials. From then on the agent works as you: your account, your organisation, your tools.
Use a dedicated account
A key opens everything its account can reach, to everyone who can talk to that agent. Create an account that holds only what the job needs, rather than using your own.
Take it back any time
Delete the key from the same page. It stops working immediately, and nothing else about your account changes.
The step-by-step for Claude in Slack is in the help centre: Connect Tulina to Claude Tag in Slack.
04
What stays private
Your passwords and keys. The keys of the tools you connect stay inside Tulina, in an encrypted vault. The agent asks Tulina to do something, Tulina does it, and the agent gets the result. It never handles the key.
Anything beyond your own access. The agent can do what you can do in Tulina, and no more. An administrator decides which tools your organisation can turn on, and access follows roles and teams.
Nothing goes unrecorded. Every piece of work the agent does is logged. Where your data lives, who runs the platform and what a security review can ask for are on the trust page.