# Can managers use 7shifts in ChatGPT for schedules?

> 7shifts in ChatGPT lets area managers check shifts and time punches in chat. The access token is unscoped, so the rules on wages and edits live in Elaichi.

**TL;DR** Yes. 7shifts in ChatGPT runs through a connector that calls the 7shifts API with one access token, and 7shifts says that token is admin-level and cannot be scoped. In Elaichi, the operations role gets an allow rule naming the schedule, punch and staff reads it needs, and block rules keep punch edits, shift deletes and wage writes out of reach. Wage reads are a separate choice, because hourly_wage also comes back on shift and punch records.

## The labor questions a restaurant group asks every Monday

Priya Shah runs operations for a restaurant group with six locations. Her area managers, including Marcus Reed, ask the same things each week. Who clocked in late at the Elm Street kitchen? Which shifts went unfilled on Saturday? How many punches are still unapproved before payroll closes? Every answer sits in 7shifts, one location at a time. 7shifts in ChatGPT puts those reads in one chat, next to the email and the sales report Marcus already has open.

MCP (Model Context Protocol) is the standard way an AI assistant calls tools in other apps. [Elaichi's `7shifts` connector](/connectors/7shifts/) calls the 7shifts API for ChatGPT, Claude, Cursor and Microsoft Copilot through one address. It signs in with an API key or with OAuth client credentials. 7shifts issues OAuth clients only to vetted partners, so a restaurant group uses an access token ([developers.7shifts.com](https://developers.7shifts.com/docs/authentication), checked November 2026).

The connector also edits punches, deletes shifts and writes wages. This playbook covers which parts a manager gets, and how wage and punch data is kept in check.

## What a 7shifts access token can reach

A 7shifts access token reaches the whole company, and nothing in 7shifts can narrow it. The authentication reference says access tokens give admin-level access to a single company. It adds that their permissions "cannot be restricted or scoped to specific resources or actions" ([developers.7shifts.com](https://developers.7shifts.com/reference/authentication), checked November 2026). 7shifts' introduction defines a company as the customer's account and a location as one restaurant ([developers.7shifts.com](https://developers.7shifts.com/docs/getting-started), checked November 2026). So one token sees every location.

Any company admin creates a token under Company Settings, then Developer Tools, then Create Access Token. The technical contact must be an active admin who has logged in at least once. Tokens are long-lived. A deleted token stops working at once, and a token idle for more than six months is deleted automatically.

Give the operations connection its own token, so it can be deleted without breaking a payroll export. Because 7shifts cannot scope the token, the rules in Elaichi do all of the narrowing.

## Which 7shifts tools the operations role should get

An operations role needs reads, and only a few. `list_all_7_shifts_shifts` filters shifts by location, department, role, user and date range. `list_all_7_shifts_time_punches` filters punches by approval status and by the flags `late`, `auto_clocked_out` and `missed_break`. `list_all_7_shifts_departments`, `list_all_7_shifts_users` and `list_all_7_shifts_contacts` cover who works where and how to reach them.

| Kind | Example tools | For the operations role |
|---|---|---|
| Schedule and punch reads | `list_all_7_shifts_shifts`, `list_all_7_shifts_time_punches` | Allow |
| Staff and structure reads | `list_all_7_shifts_users`, `list_all_7_shifts_departments`, `list_all_7_shifts_contacts` | Allow |
| Wage reads | `list_all_7_shifts_user_wages`, `7_shifts_analytics_get_hours_and_wages` | Decide per role |
| Punch and shift writes | `update_a_7_shifts_time_punch_by_id`, `update_a_7_shifts_shift_by_id` | Restrict |
| Wage writes | `create_a_7_shifts_user_wage` | Restrict |
| Deletes | `delete_a_7_shifts_time_punch_by_id`, `delete_a_7_shifts_shift_by_id`, `7_shifts_users_deactivate` | Restrict |

`list_all_7_shifts_contacts` returns mobile and home phone numbers and email for the whole company. That is useful for calling a no-show, and it is still personal data. Leave it out of a role that does not call staff.

## Where wage data shows up in shift and punch reads

Wage data is not confined to the wage tools. Each record from `list_all_7_shifts_time_punches` carries `hourly_wage` and `tips`. Each shift from `list_all_7_shifts_shifts` carries `hourly_wage` in cents. The response schema of `list_all_7_shifts_users` includes `current_wages` and `upcoming_wages`.

An Elaichi rule decides which connectors and which individual tools a target may reach. It names tools, not fields, so it cannot strip `hourly_wage` from a punch. 7shifts' own MCP page gives the same warning: with a token, an assistant "can read live wage data, employee records, and time punches" ([developers.7shifts.com](https://developers.7shifts.com/docs/mcp-server), checked November 2026).

So make the wage decision when you allow the schedule reads, not later. If an area manager already sees wages in 7shifts, the reads match their job. If not, keep the schedule and punch reads for a role that may see pay. The [BambooHR salary playbook](/blog/hr-team-bamboohr-salary-data/) works through the same choice for HR data.

## Why punch edits and shift deletes stay out of chat

7shifts treats these writes as permanent, and so should the rules. Its MCP server runs `GET` requests only. 7shifts explains that a deleted time punch, a published schedule or a modified wage record "affects real staff and real payroll, and there's no undo" ([developers.7shifts.com](https://developers.7shifts.com/docs/mcp-server), checked November 2026). Its Build with AI page adds that outside that server, an access token is admin-level and unscoped ([developers.7shifts.com](https://developers.7shifts.com/docs/build-with-ai), checked November 2026).

Punch updates carry a second trap. 7shifts' Update Time Punch reference says breaks must always be included in their correct state. Leaving out an existing break deletes it ([developers.7shifts.com](https://developers.7shifts.com/reference/puttimepunch), checked November 2026). Elaichi's `update_a_7_shifts_time_punch_by_id` repeats that warning in its own description. A model that sends only the new clock-out time can remove a recorded break. Keep punch edits in the 7shifts app.

## Writing the allow and block rules for operations

One allow rule on the operations role gives managers the reads and nothing else from 7shifts. A restriction is a rule naming which connectors and which individual tools a role or a member may reach. In Elaichi, restriction targets are role or user only; the organization default is the absence of a rule, which means allow everything. So the operations team gets its own role, and each manager holds exactly one.

Name the read tools the team uses. The same rule, or a second allow rule on the role, must also name every other app the team uses, as whole connectors. In Elaichi, once a role carries any allow rule, every connector and tool none of its allow rules names is blocked for that role. A POS or email connector left off the list is lost to the role. Do not name the 7shifts connector whole and some of its tools in one rule. Elaichi refuses that overlap on a new rule.

Then add block rules on the punch and shift writes, `create_a_7_shifts_user_wage` and the delete tools. Within a layer, a block always beats an allow. If someone later widens the allow rule to all of 7shifts, those tools stay restricted. A restriction change takes effect within about two minutes. [Per-tool versus per-app restrictions](/blog/per-tool-vs-per-app-restrictions/) compares the two shapes.

## Keeping an area manager to their own restaurants

Rules decide which tools a manager may call, not which location a call reads. For that, use a toolbox, a saved set of tools that can be shared, with a frozen parameter. A frozen parameter locks an argument on the calls that run through a toolbox entry. The AI client can see that it is set, but cannot change it.

Pin `list_all_7_shifts_shifts` and `list_all_7_shifts_time_punches` into a toolbox for Marcus's restaurant, with `location_id` frozen to that location. In Elaichi, a freeze binds only calls made through its toolbox entry, so share the toolbox with the people it governs, not the connection itself. The connection's owner keeps the 7shifts connection unshared and shares each area's toolbox at use instead.

Some reads have no location filter. `list_all_7_shifts_contacts` returns the whole company, so leave it out of an area toolbox if that matters. Do not close the unfrozen path with a restriction, because a block also applies inside the toolbox.

## Setting up 7shifts in ChatGPT and testing it

Every organization uses one address: `https://api.elaichi.ai/mcp`. Custom MCP connections with full MCP support are for ChatGPT Business and Enterprise/Edu. On Business, only admins and owners create and publish them for the workspace. On Enterprise, Configure Actions picks the actions the app may use and Configure Access picks the groups ([help.openai.com](https://help.openai.com/en/articles/12584461-connecting-custom-mcp-servers-in-chatgpt), checked October 2026). Each manager then signs in with their own OAuth grant, the sign-in that ties their Elaichi identity to ChatGPT. [Connecting Elaichi to ChatGPT](/blog/connect-elaichi-to-chatgpt/) has the full setup.

Do not rely on ChatGPT's own action settings for this. In Elaichi, connected tools are never listed one by one, however few there are. The model finds a connected tool with `search_tools` and runs it with `read_tool` (a read) or `execute_tool` (anything). ChatGPT sees only those few tool names, so its settings cannot tell 7shifts from any other connected app. [Making ChatGPT read-only in company tools](/blog/make-chatgpt-read-only-in-company-tools/) explains why the role's rules are the control.

To test, sign in as a manager on the operations role. Ask for last week's late punches at one location, then search for "time punch" and "wage". A restricted tool is left out of the tools Elaichi offers a client and cannot be called. Search still names it, flagged restricted, with no schema. Elaichi writes one entry for each connected-tool call that reaches execution (a call refused earlier writes none). Each entry records the tool, the outcome and the client, plus the one path argument that names the object, as the target id, and nothing else about the arguments.

Mind the limits on both sides. 7shifts' Build with AI page puts its API limit at 10 requests per second per IP address ([developers.7shifts.com](https://developers.7shifts.com/docs/build-with-ai), checked November 2026). Its errors reference says a client past 600 requests a minute is blocked for one minute ([developers.7shifts.com](https://developers.7shifts.com/reference/errors), checked November 2026). Elaichi allows 120 MCP requests a minute for each OAuth grant. When a manager needs a restricted tool, they file an access request. An admin approves it under Governance, then Access requests, and the grant lifts that one tool for that one person.

## When 7shifts' own MCP server is enough

For one developer or analyst, 7shifts' own MCP server may be all that is needed. It is public at `https://developers.7shifts.com/mcp`, and its docs cover Claude Code, Cursor, VS Code and Claude ([developers.7shifts.com](https://developers.7shifts.com/docs/mcp-server), checked November 2026). With a token in the config, it makes live GET requests. It never writes, by design. That is a sound fit for exploring the API or drafting a report script.

Elaichi earns its place when several managers ask, from ChatGPT and other clients, about 7shifts and the rest of their apps. The token is entered once in Elaichi, not pasted into each manager's config. One set of rules decides who sees wages, and one audit trail shows who asked what. If one person only wants to test a few queries, [you may not need a gateway yet](/blog/when-you-dont-need-an-mcp-gateway/). Browse the [900+ connectors](/connectors/) for the rest of the stack.

## FAQ

### Does 7shifts have its own MCP server?

Yes. 7shifts documents a public MCP server at https://developers.7shifts.com/mcp that reads its API spec and guides, and can make live GET requests when an access token is added to the client's config. 7shifts says it exposes GET endpoints only, so nothing can be changed or deleted through it (checked November 2026).

### Can a 7shifts access token be limited to read-only or to one location?

No. 7shifts' authentication reference says access tokens give admin-level access to a single company, and their permissions cannot be restricted or scoped to specific resources or actions (checked November 2026). Any narrowing has to happen in the layer that holds the token, such as Elaichi's per-role rules.

### Will ChatGPT see employee wages through 7shifts?

Only if the role can call a tool that returns them. In Elaichi's 7shifts connector, list_all_7_shifts_user_wages and 7_shifts_analytics_get_hours_and_wages return pay, and the shift and time punch reads also carry an hourly_wage field. A rule in Elaichi names tools, not fields, so decide on wages before you allow those reads.

### Can ChatGPT edit or delete 7shifts time punches?

Through a connector with an unscoped token, it could. Elaichi's 7shifts connector has update_a_7_shifts_time_punch_by_id and delete_a_7_shifts_time_punch_by_id. A block rule on each for the operations role keeps them out of reach. The tools are withheld from ChatGPT, and search names them only as restricted.

## Read next

- [Keep BambooHR salary data out of AI assistants](/blog/hr-team-bamboohr-salary-data/) — Two layers keep BambooHR salary data away from Claude and ChatGPT: the access level behind the API key, and an allow rule per role in Elaichi.
- [Read-only ChatGPT access to your company apps](/blog/make-chatgpt-read-only-in-company-tools/) — Read-only ChatGPT access comes from a restriction rule on a role, not from a consent checkbox or a client setting that sounds like one.
- [Connect Elaichi to ChatGPT](/blog/connect-elaichi-to-chatgpt/) — To connect Elaichi to ChatGPT, a workspace admin adds one MCP endpoint as a custom app and each person signs in with OAuth. No API key, no per-user URL.
