Perplexity custom connectors: add a remote MCP tool
Check your organization’s permission, prepare the remote server settings, and test a custom connector with a small request.
Updated 5 min read
Perplexity custom connectors let it call an outside MCP server, a service that gives an AI app tools it can use. The verified app setup is Enterprise and administrator-controlled (opens in a new tab), so confirm access before looking for an Add control. A connection to fenbs can then keep the work and the next step outside an individual conversation.
What do you need before adding a connector?
- Ask your organization’s administrator whether members may add custom connectors. Do not assume your subscription alone gives you this permission.
- Get the remote MCP Server URL from the service you want to connect. It must use HTTPS.
- Confirm the server’s supported Authentication and Transport with its operator before filling in the form.
- Decide whether this is your own connection or an organization-wide one. Have an administrator handle organization sharing.
- Choose one harmless reading task that will show whether the connection works.
Prepare the purpose of the connection as carefully as the address. For example, “read the current delivery checklist” is a useful first goal. “Give the assistant access to everything” does not tell you what success looks like or which access would be excessive.
Use the vendor’s current setup guide to open the custom connector form. Once there, use the fields it documents rather than treating an older screenshot as proof that your account has the feature. If the form is unavailable, resolve eligibility with your administrator first.
How do you fill in the remote connector form?
The remote connector guide (opens in a new tab) documents Name, MCP Server URL, Authentication, and Transport. Give the connector a name that tells you which service and account it reaches. A name like “Studio checklist, read access” is easier to review later than “New connector.”
- Enter Name and the HTTPS MCP Server URL supplied by the service.
- Choose the Authentication the server requires: OAuth, API Key, or None. Do not choose None to work around a failed sign-in.
- Match Transport to the server: Streamable HTTP or SSE. These are two supported ways to carry MCP messages over the network.
- Read and acknowledge the custom-connector risk statement, then choose Add.
- Open the connector card and complete the required authentication and enabling steps before testing it.
OAuth means signing in and granting access through the service. Where that server does not support dynamic client registration, the guide says a client ID and secret may be needed. Get those from the approved service setup; never put secrets into an ordinary prompt or a shared troubleshooting note.
Keep a plain record of the intended account, the person responsible for access, and the first task to test. How fenbs works explains how connected assistants share the work around a connection without making the connection settings themselves your lasting work record.
How do you check that the tools can actually work?
Start with a question whose answer you can inspect in the original service. Ask the assistant to read one named record and report a small detail. Compare that result with the source before asking it to change anything.
- Correct result: confirm that it came from the intended service and account, then keep the successful test.
- No result: check the address, authentication completion, and server transport before rewriting the whole prompt.
- Permission refusal: ask which access is missing and whether your role should have it. A connection is not permission for every action.
- Unexpected data: stop and review the account or organization scope before making another request.
For a first write test, use a clearly named sample record that you are allowed to change. State exactly what may change, ask for review before an important action, and inspect the actual record afterward. If you cannot verify the result, mark the test incomplete.
Use Perplexity prompt patterns for the wording of a small test. Do not confuse a vague request with a broken connection: ask for one source, one result, and one limit on what the assistant may do.
Is this the same as a file connector or Perplexity MCP?
No. A file-app connector (opens in a new tab) supplies files to research. Its documented workflow includes authorization, a Connected state, and the … → Upload control. Those instructions do not establish that every Pro account can add an arbitrary MCP tool server.
| What you want | Which direction to check |
|---|---|
| Perplexity calls an outside tool | A custom MCP connector inside Perplexity. |
| Another assistant uses Perplexity research | Perplexity’s own MCP server, covered separately. |
| A program uses a saved remote connection | The Agent API connector workflow in the API Console. |
| Research uses a file from a connected app | The file-app connector workflow. |
The Perplexity MCP article covers tools flowing in the other direction. Developers should use the separate Agent API connector documentation (opens in a new tab); its Project connectors → Add custom connector workflow is not a consumer-app menu.
What about local servers and shared access?
This procedure covers the verified remote HTTPS route. The cited setup facts do not provide a verified local installation command, so do not substitute a localhost address or paste an unexplained command into your computer. Have the server operator confirm the supported route for the service you are using.
The custom connector guide (opens in a new tab) distinguishes a private individual connector from an organization connector. Organization OAuth can use shared administrator authorization or individual member authorization. Ask which model applies before assuming that another person will see the same data you do.
Write down who owns the connection and who should review access changes. If you are also organizing reusable research, Perplexity Spaces explained is a separate guide. Putting information in a Space is not the same test as authorizing an outside tool.
Never lose the work you did in an AI session
After a useful AI session, save it in fenbs: the key observations, findings, research and context. Then carry on from exactly where you left off, on any device and in any other AI. The next AI you connect reads it at the start of the session, so you never have to explain the same problem again.
Create a free accountFree for up to 5 people and 50 open tasks · No credit card · Works with Claude, ChatGPT, Cursor and more
How to keep the work behind a connection with fenbs
- Create a free account.
- Save what you are trying to do, the rules to follow, and the next small task. Keep passwords and tokens out.
- Connect Perplexity to fenbs if your organization allows custom connectors; start with the connections page.
- Ask the connected AI to read where you left off before working, then save what changed before it stops. You can carry on from there on another device or in another AI.
fenbs keeps the shared work; it does not grant Perplexity connector eligibility or bypass an administrator. For one disposable file lookup, a lasting work record may be more than you need. Browse connecting AI tools when you need to understand the connection before choosing one.