Connect an AI assistant
Ask the restaurant owner or an authorised controller to connect the assistant.
Use an AI client that supports TablePort's remote OAuth MCP connection.
Connect
- In the AI client's supported connector settings, add TablePort or a custom
remote MCP connection. Client menu names differ by product and release.
- For the production connection, use
https://apis.tableport.io/v2/mcp.
Development and staging use their own configured endpoints; do not reuse
credentials across environments.
- Complete the TablePort sign-in and authorisation flow. Check the requesting
client, account, business context, and permissions.
- Review the requested
mcp:business:read and, when requested,
mcp:business:write scopes. A write scope does not grant every possible
product action.
- Return to the client and inspect its available tools. Ask the assistant to
identify the intended location from authorised records before using it.
Do not add private credentials or a business identifier to the connection URL.
If your client does not provide a compatible connection flow, ask support for
the supported setup for that client.
Check the connection
Ask for a small read, such as the locations or menus available to this account.
Confirm that the response concerns your restaurant and expected location. Do not
test a new connection by taking a payment or changing guest data.
If it does not connect
- Sign-in or authorisation fails: check the TablePort account and the
client's supported OAuth flow.
- No tools appear: recheck the environment, granted permissions, and
account. An unavailable feature or restricted catalogue may also explain it.
- A particular action is unavailable: compare it with
supported external capabilities. Reconnecting does not
add unsupported actions.
- Authorisation has expired: use the client's normal reauthorisation flow.
Do not paste a token into chat.
Related: Agent recovery and
changes and approvals.