The Nest · news
Crow Nest is now in the Claude Connectors Directory
Crow Nest is now listed in the Claude Connectors Directory. To add it, open Customize > Connectors in Claude, select Discover, search for "Crow Nest" and select Connect to Claude. Sign in once, and Claude can publish the apps it builds and share them with your team behind sign-in, without leaving the chat.
Until now you had to add Crow Nest as a custom connector by pasting our server address into Claude. That still works. The directory listing just means you can find us the same way you find any other connector, and your company's Claude owner can turn it on for everyone at once.
If you haven't met us yet: Crow Nest is where you put the apps you build with Claude, ChatGPT, Codex or Cursor so your team can use them, behind your company's sign-in, with a database built in, and visible to IT. The rest of this post covers how to connect it, what you can ask Claude to do once you have, and when you don't need us at all.
#What changed
Connectors are how Claude reaches outside services. Anthropic's guide to getting started with connectors describes two ways to add one: pick it from the directory, or add a server by its URL as a custom connector.
Before today, Crow Nest was only available the second way. That meant copying https://crow-nest.com/mcp, finding the custom connector form, and naming it yourself. It worked fine, but it's a step most people only take when someone tells them to. And on Claude's Free plan, a custom connector uses up the one custom slot Anthropic allows, according to Anthropic's help article on the connectors directory.
Now Crow Nest has its own listing page in the directory, with a description of what it does, a link to our docs and privacy policy, and a Connect to Claude button. Nothing about the server itself changed. If you already connected Crow Nest as a custom connector, you don't need to do anything; it's the same server at the same address.
#What the Community label means
Our listing carries a Community label, and we think you should know what that means before you connect. According to Anthropic's page on connector verification, every submitted connector is scanned automatically for policy compliance and listed as Community by default. Anthropic screens Community connectors but hasn't reviewed them in depth, and it may later move connectors to a Verified review where its reviewers test each tool by hand. There's no application for that; it happens on its own or it doesn't.
So the label is a statement about Anthropic's review, not a grade. The same page says it plainly: only connect to developers you trust. We'd rather you read how Crow Nest handles your data in the security docs than take the badge on faith.
#How to connect Crow Nest to Claude
On your own account (any plan):
- Go to Customize > Connectors on claude.ai, or select Customize in the sidebar of the Claude desktop app, then Connectors.
- Select Discover, type "Crow Nest" into Search connectors, and open the listing.
- Select Connect to Claude.
- A Crow Nest sign-in page opens. Sign in with Google, Microsoft or an email code, then approve the connection.
- Back in Claude, Crow Nest shows under Your connectors as Connected.
If you don't have a Crow Nest account yet, you make one in step 4. There's nothing to set up beforehand.
On a Team or Enterprise plan: an Owner adds connectors for the whole organization. If you're an Owner, the button on the listing reads Connect for your team, which adds Crow Nest for everyone; you then connect your own account. If you're a member and it hasn't been added yet, you'll see Request instead, which asks an Owner to add it. Adding it doesn't sign anyone in; each person still connects their own Crow Nest account.
In Claude Code: Anthropic's connectors guide says the connectors you add in claude.ai are available in Claude Code when you sign in with your Claude account. So connecting once in Claude covers both. If you prefer to add it from the terminal, the command is in the MCP server docs, along with steps for ChatGPT, Codex, Cursor, VS Code, Windsurf and Gemini CLI.
In a conversation: select + in the message box, then Connectors, and make sure Crow Nest's toggle is on. The toggle works per conversation. Claude may ask before using a tool; choose Allow once or Always allow.
#What you can ask Claude to do
Once Crow Nest is connected, you don't need to learn any commands. You ask for what you want, the way you'd ask a coworker. Here are requests that work, with the Crow Nest tool Claude uses for each (the full list is in the MCP server docs):
| You ask Claude | What happens |
|---|---|
| "Make a PTO tracker for my team and publish it to Crow Nest." | Claude builds it, reads our developer reference, and publishes it. You choose the name and who can see it. |
| "Share it with sam@example.com and priya@example.com." | Adds those two people. The change reaches them within 60 seconds. |
| "How many people opened the PTO tracker this week?" | Shows opens and unique viewers for the last few days. |
| "That last change broke something. Roll it back." | Makes the previous version current again. |
| "Make it save entries so they're still there tomorrow." | Uses crow.db, the database every app gets, with rules for who can read and write. |
| "Put it on tools.mycompany.com." | Connects your own subdomain and lists the DNS records to add. |
| "Set up an organization and invite my team." | Creates an organization and sends invitations by email. |
The useful part is that this all happens in the same conversation where you built the thing. You don't export a zip, open a dashboard, or learn a deploy command. When you want a change, ask for it and say "publish again." Every publish is a new version at the same address, and older versions are kept, as the quickstart explains.
#What publishing asks you
Two questions come back to you every time Claude publishes a new app, and Crow Nest never answers them on your behalf.
A name. It becomes the app's address, like pto-tracker.crow-nest.com.
Who can see it. The five choices are in the sharing docs:
- Private: only you.
- Specific people: you and up to 200 email addresses you add.
- Link: anyone who has the link and signs in.
- Organization: everyone in your organization or company.
- Public: anyone on the internet, no sign-in.
When your AI app supports it, the question comes to you directly as a form the AI can't read or fill in. Otherwise Claude asks you in the chat and waits for your answer. Your teammates then open the link and sign in with their work Google or Microsoft account, or an email code. They don't need a Claude account, a Crow Nest plan, or a password you made up.
If the app saves data, ask Claude to use crow.db. Each app gets its own database, and the rules for who can read and change each collection are enforced on the server, not in the browser. "Everyone can see requests, only the person who made one can edit them" is a rule you set, not code you write. The data docs show how.
#Claude artifacts or Crow Nest?
Claude can already share what it builds as an artifact, and Claude Code can publish artifacts too. So it's fair to ask why you'd connect anything at all. Sometimes you shouldn't.
According to Anthropic's help article on sharing artifacts, on Team and Enterprise plans artifacts stay inside your organization by default and everyone who opens one needs a Claude account. When Anthropic brought artifacts to Claude Code in June, it said those artifacts are viewable only by signed-in members of your organization, cannot be made public, and keep a version history.
An artifact is the better choice when:
- everyone who needs the app already has a Claude seat in your organization,
- the app is small and doesn't need its own address or domain,
- you don't need server code, schedules, webhooks or outside API keys,
- you want the least possible setup and are happy for it to live inside Claude.
Crow Nest is the better choice when:
- people outside your Claude workspace need it: a client, a contractor, the half of the company that doesn't use Claude,
- the app saves data and you need rules about who can change what,
- it calls an outside API with a secret key, or needs server code that runs on a schedule,
- it should live at your own address,
- IT needs to see what's been built, who owns it and who can open it.
We wrote a longer walkthrough of all the options, from sending the code to general web hosts, in How to share an app you built with Claude Code with your team. If you're weighing a general host, our Crow Nest vs Vercel comparison says when Vercel is the better pick.
#What never goes through the chat
Connecting any service to an AI raises a fair question: what does the AI get to see and do? Here's how Crow Nest is set up, from the MCP server docs and the security docs:
- Secrets and AI keys never pass through the conversation. When an app needs an API key, Claude gives you a link to the Crow Nest dashboard and you paste the key there. Keys are encrypted at rest (AES-256-GCM, a separate key per app) and never sent to the browser.
- Card details are only typed on Stripe's pages. Claude can tell you your plan and usage, and links you to Stripe when you need to act.
- Publishing refuses files that contain API keys or tokens, so a key Claude left in the code doesn't go live by accident.
- Every tool says whether it reads, changes or deletes something. That lets Claude ask before changes, and you can set each tool to Always allow, Needs approval or Blocked in the connector's settings in Claude.
- Deleting an app needs your confirmation, and Claude only does it when you ask.
- You can cut Claude off any time, either from Claude's connector settings or from the Tokens page in the Crow Nest dashboard.
#What it costs
Connecting is free, and so is browsing the directory. On our side, publishing and hosting are free. You add a card when you sign up: on Pay as you go nothing is charged until your apps use paid services like storage and server code, and the first $1 of usage each month is free. Crow Nest Personal is a flat $7 a month, and companies we set up pay $5 per person per month. People you share an app with pay nothing to open it. The numbers are on the pricing page.
If you want to see a published app before you connect anything, try the live demo.
#Frequently asked questions
#How do I add the Crow Nest connector to Claude?
Open Customize, then Connectors, in claude.ai or the Claude desktop app. Select Discover, search for "Crow Nest", open the listing and select Connect to Claude. A Crow Nest sign-in page opens, where you sign in with Google, Microsoft or an email code and approve the connection. On Team and Enterprise plans an Owner adds it for the organization first, then each person connects their own account.
#Do my teammates need Claude to open an app I publish?
No. An app on Crow Nest is a regular web app at its own address. People you share it with sign in with their work Google or Microsoft account or an email code, and open it in any browser. Only the person publishing needs Claude, or another AI connected to Crow Nest. This is the main difference from a Claude artifact, which every viewer needs a Claude account to open.
#What does the Community label on the listing mean?
It means Anthropic scanned Crow Nest automatically before listing it but hasn't done its in-depth, hand-tested review, which it calls Verified. Every new submission starts as Community, and Anthropic decides on its own which ones to escalate. The label doesn't change what the connector can do once connected. Anthropic advises connecting only to developers you trust, so read our security docs first.
#Does the connector work in Claude Code?
Yes. Anthropic says connectors you add in claude.ai are available in Claude Code when you sign in with your Claude account, in the terminal and in cloud sessions. You can also add Crow Nest from the terminal with one command, listed in our MCP server docs, which is handy if you use Claude Code with an API key instead of a Claude login.
#Can Claude see my API keys or card details through the connector?
No. When an app needs a secret or an AI key, Crow Nest gives Claude a link to the dashboard, and you paste the value there, so it never enters the chat. Card details are only typed on Stripe's pages. Keys are encrypted at rest with a separate key per app and are never sent to the browser. Publishing also refuses files that contain keys.