Connect to your own AI

You already pay for an AI. Use it.

TribeNest speaks MCP (Model Context Protocol), the open standard for giving an AI assistant real tools instead of guesses. Connect it once and Claude or ChatGPT can build your website, publish it, create events with ticket tiers, upload media, write blog posts, and read your real numbers, all from an ordinary chat.

Pick your assistant

Both work the same way and get the same tools. Jump to yours:

Everything after the two setup sections applies to both: what it can do, how to get great results, safety and limits, and troubleshooting.

The one thing to copy

Whichever assistant you use, this is the only address you need:

https://api.tribenest.co/mcp

You'll paste it once. There's no API key to create, nothing to install, and no terminal. TribeNest asks you to sign in and pick a profile, then hands your assistant a key that only works for that profile.

Connect Claude

Custom connectors are available on Claude's paid plans (Pro, Max, Team, and Enterprise). You can connect from the web app, the desktop app, or mobile. The connection follows your account, so doing it once covers all of them.

1. Open Claude and go to Settings.
2. Choose Connectors (on some plans it sits under Connectors → Browse / Add).
3. Click Add custom connector.
4. Give it a name (TribeNest) and paste the URL:

https://api.tribenest.co/mcp

5. Click Add / Connect. Claude opens a TribeNest window.
6. Sign in to TribeNest if you're not already, then on the authorization screen:
   • Read what the connector will be able to do.
   • Choose the profile you want it to manage.
   • Click Authorize.
7. Claude closes the window and shows TribeNest as connected. In a new chat, make sure the TribeNest connector is enabled (Claude shows connectors in the chat's tools menu), and you're ready.

Test it with something harmless:

What's my TribeNest profile, and what can you do for it?

If it answers with your profile's name, you're connected.

Connect ChatGPT

ChatGPT supports custom MCP connectors on its paid plans (Plus, Pro, Business, and Enterprise). On personal plans, custom connectors live behind a developer / advanced setting you turn on once.

1. Open ChatGPT → Settings.
2. Go to Connectors (also shown as Apps & connectors in some versions).
3. If you don't see an option to add your own, open Advanced settings and enable developer mode. That's what exposes custom MCP connectors.
4. Choose Create / Add custom connector and fill in:
   • Name: TribeNest
   • MCP server URL: https://api.tribenest.co/mcp
   • Authentication: OAuth (not "no authentication" and not an API key)
5. Save. ChatGPT opens a TribeNest window.
6. Sign in, pick the profile to connect, and click Authorize.
7. Back in ChatGPT, start a new chat and enable the TribeNest connector for it (in the message composer's tools / connector picker).

Test it:

Using TribeNest, tell me which profile you're connected to.

What happens when you authorize

Worth understanding, because it's what keeps this safe:

• One profile per connection. The key your assistant receives works for the single profile you picked and nothing else. Other profiles on your account stay invisible to it.
• Your permissions, not more. The connection can do exactly what you can do on that profile. If you're a team member without store access, neither is your assistant.
• No passwords are shared. You sign in on TribeNest's own screen. Your assistant never sees your password.
• Revocable in one click, any time. In TribeNest go to Account settings → Connected apps, find the assistant, and click Disconnect. It stops working immediately.
• Every action is logged. Tool calls are recorded in your profile's audit trail, so you can always see what was done by chat.

To connect a second profile, repeat the setup and pick the other profile. You'll have two connections listed under Connected apps.

What your assistant can do

Once connected, your assistant isn't guessing about your business. It's reading and writing the real thing.

Build and run your website

• Build a custom-code website end to end: create files, edit them, start a live preview, check for errors, and publish.
• Update editable text and images on an existing site (changes go live without a rebuild).
• Roll back to an earlier snapshot, list versions, and update the site's theme.
• Manage page actions (the buttons and calls-to-action on your pages).

Sell

• Create events with their ticket tiers in one step, which is ideal for moving an event over from somewhere else.
• Create membership tiers, offers, coupons, and donation setups.
• Look up your real product, event, and course IDs so links and buttons point at the right thing.
• Read orders, refunds, expenses, transactions, and a profit-and-loss summary.

Publish and grow

• Write and publish blog posts, members-only posts, and podcast episodes.
• Create forms, collections, and collection entries, and query them.
• Upload media from a link or file, then use those real URLs on your site.
• Schedule and publish social posts, read comments and your inbox, and check social analytics.
• List email campaigns, smart links, reviews, community posts, and moderate what needs moderating.

Operate

• Create projects and tasks, log time, and turn tracked time into an invoice.
• Build and test automations.
• Create and manage apps on your site, including their secrets and permissions.

Getting great results

The difference between a frustrating session and a great one is almost entirely in how you ask.

Start with the outcome, not the mechanics. "I want a homepage that leads with my new EP and a store page where fans can buy the vinyl" beats "create index.tsx". The assistant knows the mechanics; only you know the outcome.

Let it look before it leaps. For anything beyond a small edit, ask it to inspect first:

Look at my current site and my products, then tell me what you'd change on the homepage before changing anything.

Work in small, publishable steps. One page or one section per request. Publish, look at it, then continue. Long unbroken build sessions are harder to review and harder to undo.

Ask for a preview before publishing. "Start a preview and give me the link" lets you see the change on a real URL before your visitors do.

Insist on real data. Never let it invent an image URL, product id, or price. If something looks made up, say: "Use my real media library and product ids, and don't invent any."

Ask for errors to be checked. "Check for errors and fix anything you find, then publish." The connector has a real error check; use it before every publish.

Name the profile when you have more than one. If you've connected two profiles in two chats, start with "Work on the pinelight profile" to remove all doubt.

Keep one chat per project. A chat that built your store page carries all that context. Starting a fresh chat for an unrelated task keeps both sharper.

Good opening lines to steal:

Build me a site: a bold homepage featuring my latest release, a store page, and an email signup. Use my real media.

Add a tour section to my homepage that lists my upcoming events, then start a preview.

Create an event for August 14 at The Shrine with three ticket tiers: early bird £15, standard £25, VIP £60.

My homepage headline should read "Live in Lagos this August". Publish that text change.

What sold this month, and what's my profit after expenses?

List the TribeNest actions you can perform for me, grouped by what they're for.

Safety and what it cannot do

• It can't touch another profile, or another creator's anything. One connection, one profile.
• It can't exceed your own permissions on that profile.
• It can't create store products. Products are made in the dashboard (they carry files, variants, and pricing that deserve a real form). Your assistant can read them, link to them, and put them on your site, and it can create events, ticket tiers, membership tiers, offers and coupons directly.
• It can't take payments or move money. It reads financial records; it doesn't process refunds or payouts on your behalf without you doing so in the dashboard.
• It can't change your account, billing, or plan.
• Publishing is a deliberate act. It publishes when asked. If you'd rather see everything first, tell it up front: "Never publish without showing me a preview and asking."
• Rate limits apply. Very long bursts of tool calls get throttled. If your assistant reports being rate limited, wait a minute and continue.

Which path should I use?

You can move between all three freely: it's the same site, the same versions, the same content. Nothing is locked to the way you started.

Troubleshooting

• The assistant says it has no TribeNest tools. The connector is off for that chat. Enable it in the chat's tools/connector picker, then ask again.
• "I can't find that product / event." It's probably connected to a different profile than the one you're thinking of. Ask "Which profile are you connected to?" and re-authorize with the right one if needed.
• It's writing code that doesn't work. Tell it: "Call read_build_guide_first, then check_errors, and fix what you find." Nearly every broken build comes from an assistant that skipped the guide.
• It won't add the connector at all. Custom connectors need a paid plan on both Claude and ChatGPT, and on ChatGPT they may need developer mode enabled first.
• Changes aren't showing on my live site. Content edits are instant; code changes need a publish. Ask "Publish my site" and then reload.
• You want to stop everything, now. Account settings → Connected apps → Disconnect. Access dies with the click.