Skip to content

Connect your agent

Connect Senvi once, then start sessions in the agent you already use. Pick your agent, install with one click or two commands, approve in the browser, paste one line. You do not need an API key, and you do not need an account before you start: the approval page makes one.

Connecting is free, and so is getting set up (setup prompt and onboarding intake). The weekly loop (session) needs a plan, or the 7-day trial that starts when you sign up.

Set up Senvi here

Which agent do you use?

  1. Install

    Choose your agent and this step shows what to click.

    Or let your agent do this step

    Copy the setup prompt and paste it into your agent. It connects Senvi and guides you through any steps it needs you to take.

  2. Approve the connection

    Your agent opens Senvi’s approval page in your browser. Sign in there, or make your account there, and approve.

    Sign in to see whether your agent is connected.

  3. Start your session

    Paste this line into your agent.

    Start my Senvi session: call list_skills to confirm you are connected, then fetch get_skill("setup-prompt") and follow it.

Steps by agent

The same setup, written out for each agent, if you would rather connect by hand or your agent gets stuck partway. Each tab ends with a prompt you can paste to have the agent run those steps for you.

Claude Code

We have not connected from this client ourselves yet. The steps below come from the vendor's own documentation.

Who this is for. You run Claude Code in a terminal.

Where the setting lives. Claude Code adds a remote server from the terminal and keeps it in its own user config, so Senvi is there in every project you open. You never edit that file.

Steps. Three, and only the second one leaves the terminal.

  1. Run the first command below. Claude Code adds Senvi to its server list.
  2. Run the second. Claude Code signs in with a login command, which opens a browser window on Senvi's approval page. Approve it there.
  3. Start a new Claude Code session. Senvi's tools arrive with it.

The two commands:

claude mcp add --transport http --scope user senvi https://mcp.senvi.ai/api/mcp
claude mcp login senvi

If the second command is one your Claude Code does not know, skip it: start a new session, type /mcp, pick Senvi, and approve in the browser it opens.

What the approval looks like. A Senvi page on senvi.ai, asking you to allow the connection and naming the account you are signed in as. It is our page, not a third party's, and nothing on it asks you for a key.

How to tell it worked. /mcp in the new session lists Senvi as connected, and the connection line at the top of this page flips on its own.

What to say to start. start my Senvi session

There is nothing else to set up. Claude Code speaks streamable HTTP and OAuth, and Senvi hands out its own credentials to a client that asks, so there is no key to copy and no field to fill in.

The Claude apps

We have not connected from this client ourselves yet. The steps below come from the vendor's own documentation.

Who this is for. You use Claude on the web, desktop, or mobile. You can also paste the guided setup prompt into Claude and follow its instructions there.

Where the setting lives. On the web or desktop, open Customize, then Connectors.

Steps. Connect your Senvi account once.

  1. In Claude on the web or desktop, open Customize, then Connectors, and choose the option to add a custom connector.
  2. Enter Senvi as the name and paste the server URL below.
  3. Choose Connect, then sign in to Senvi and approve the connection in the browser window.
  4. Enable Senvi in the conversation's connectors or tool list.
https://mcp.senvi.ai/api/mcp

What the approval looks like. A Senvi page opens in your browser and asks you to allow the connection.

How to tell it worked. Ask Claude to check the Senvi connection with list_skills. A successful response confirms that Claude can reach your signed-in Senvi account. Then continue with the guided setup prompt.

What to say to start. start my Senvi session

If the tools do not appear, check that Senvi is enabled for this conversation. Reconnect if sign-in expired, then repeat the check.

Custom connectors are available on Free, Pro, Max, Team and Enterprise. Free accounts can connect one custom connector at a time. For Team and Enterprise, an organization owner first adds Senvi under Organization settings, then Connectors, using Add custom web connector. Each member then connects their own Senvi account and enables it in the conversation.

The connection belongs to your Claude account and can be used on mobile too. Senvi runs remotely, so you do not need to run a local server.

Cursor

We have not connected from this client ourselves yet. The steps below come from the vendor's own documentation.

Who this is for. You work in Cursor.

Where the setting lives. Cursor's MCP list, which Add to Cursor fills in for you through Cursor's one-click install deeplink. By hand, it is a url entry in .cursor/mcp.json for one project, or the same entry in ~/.cursor/mcp.json for every project you open.

Steps. Two, and the second one happens in your browser.

  1. Click Add to Cursor. Cursor opens on the Senvi server and asks you to confirm.
  2. Click Senvi's sign-in control in that list (it says Needs login, or Connect), then approve in the browser window it sends you to.
Add to Cursor

If the button does nothing, open .cursor/mcp.json in the project, or create it, and add Senvi as a server with the address below as its url. Then open Cursor Settings, search for MCP, and click Senvi's sign-in control there.

The address:

https://mcp.senvi.ai/api/mcp

What the approval looks like. The same Senvi page in a browser window, asking you to allow the connection.

How to tell it worked. Cursor lists Senvi's tools under the server, and the connection line at the top of this page flips on its own.

What to say to start. start my Senvi session

There is nothing else to set up. Cursor speaks Streamable HTTP and OAuth, and Senvi hands out its own credentials to a client that asks, so there is no key to copy and no field to fill in.

Codex

We have not connected from this client ourselves yet. The steps below come from the vendor's own documentation.

Who this is for. You run Codex in a terminal.

Where the setting lives. Codex adds a remote server from the terminal and keeps it in ~/.codex/config.toml. You never open that file.

Steps. Three, and only the second one leaves the terminal.

  1. Run the first command below. Codex adds Senvi to its server list.
  2. Run the second. Codex signs in with a login command, which opens a browser window on Senvi's approval page. Approve it there.
  3. Start a new Codex session. Senvi's tools arrive with it.

The two commands:

codex mcp add senvi --url https://mcp.senvi.ai/api/mcp
codex mcp login senvi

What the approval looks like. A Senvi page on senvi.ai, asking you to allow the connection and naming the account you are signed in as. It is our page, not a third party's, and nothing on it asks you for a key.

How to tell it worked. /mcp in the new session lists Senvi with its tools, and the connection line at the top of this page flips on its own.

What to say to start. start my Senvi session

There is nothing else to set up. Codex speaks streamable HTTP and OAuth, and Senvi hands out its own credentials to a client that asks, so there is no key to copy and no field to fill in.

Any other MCP client

Who this is for. Anything that speaks MCP and can do OAuth, including clients we have never opened.

Where the setting lives. Wherever your client adds a remote MCP server. We have not seen that screen, so we are not going to name a menu that might not be there.

Steps. Two. There is no third one.

  1. Add Senvi as a remote MCP server over streamable HTTP, at the address below.
  2. Approve the connection in the browser window that opens.

The address:

https://mcp.senvi.ai/api/mcp

What the approval looks like. The same Senvi page in a browser window, asking you to allow the connection.

How to tell it worked. Your client lists Senvi's tools.

What to say to start. start my Senvi session

Senvi runs its own OAuth server, with open dynamic client registration and PKCE. Your client registers itself and collects its own credentials, so there is no key for us to issue and nothing for you to paste.

If a client will not register at all, the discovery documents are the place to start. They sit at /.well-known/oauth-authorization-server and /.well-known/oauth-protected-resource on the server address above.

Where the coaching comes from

Your agent fetches the coaching from Senvi at the start of each session, so it stays current and it stops when a plan stops. It arrives as ordinary text in your agent's context, with one line naming the account it was served to.

What next

The weekly loop is what you do once the first session is done.

Troubleshooting is where to go if the approval did not take, or if your agent connected and then acted like it had never heard of us.

Sources