Slack Classic Apps Deprecation Was Dated 16 November. Slack Paused It, and Migrating Still Changes Your Bot.
The cutoff moved three times, then lost its date. The reasons to migrate did not move, and neither did the behaviour changes waiting on the other side.
The Slack classic apps deprecation carries today's date in Slack's own changelog. The entry reads that from 16 November 2026 classic apps will no longer function, their API calls will be rejected, and they will stop receiving Events API events. A notice near the top of the same page says the opposite: the deprecation is paused and classic apps will continue to work.
Both statements are live on one page. If your team has a calendar reminder, a vendor migration guide or a compliance tracker that says a bot dies today, this is why, and the reminder is out of date. The more useful question is what to do with an old bot now that the forcing date is gone.
The Slack classic apps deprecation date that arrived and did nothing
Slack announced the end of classic apps on 30 September 2024, in a changelog entry that also set a date for legacy custom bots. The page has been edited in place since. Its opening paragraph now shows four successive dates in a row, and its "What's changing" section still gives 16 November 2026 as the day classic apps stop.
The statement that governs is a later one. On 8 December 2025 Slack published a short notice saying that, based on feedback from customers and developers, it had decided to pause the migration. Classic apps would continue to work for the foreseeable future. No replacement date came with it. Slack's help centre retirement list shows the retirement date for legacy classic apps as "On hold", with the status still "Announced". Slack's deprecation changelog shows no later entry on classic apps as of publication.
We have not tested a classic app against the API today, and you should not need to. If yours is still answering, that is the pause working as described.
Three postponements, then no date at all
The pause was the fourth change to the timeline in fourteen months. Each of the three before it used the same sentence about consideration and feedback.
| Announced on | Cutoff for classic apps | Slack's stated reason |
|---|---|---|
| 30 September 2024 | September 2025 | Too many ways to create an app |
| 7 November 2024 | 31 March 2026 | Consideration and feedback |
| 11 April 2025 | 25 May 2026 | Consideration and feedback |
| 29 July 2025 | 16 November 2026 | Consideration and feedback |
| 8 December 2025 | Paused, no date | Feedback from customers and developers |
One date in that original announcement did hold. Legacy custom bots, the older integration type created from the Marketplace bots page, lost support on 31 March 2025 as scheduled, and Slack said their API calls would be rejected from that day. So Slack does carry these cutoffs through. It did so for the smaller population and blinked on the larger one.
That history tells you how to read the pause. A platform that postpones three times is hearing from customers who cannot migrate in time. A platform that then removes the date has not changed its mind about the old model. It has stopped promising when.
What the pause leaves in place
Slack's wording is specific. Classic apps keep working, they are still a legacy method, new ones cannot be created, and you should migrate "to avoid any future service interruptions". Four practical constraints follow from that, and none of them was lifted in December.
- You cannot create a new classic app. Creation has been blocked since 4 June 2024. A bot you delete by accident, or a second copy you want for staging, cannot be rebuilt on the old model.
- Classic apps run on the umbrella bot scope, one permission that covers a long list of API methods. Slack's own RTM documentation notes that many workspace administrators refuse integrations with scopes that broad.
- Newer platform features are closed to them. Socket Mode, for example, is only available to apps on granular permissions.
- The pause notice commits to no minimum warning before the deprecation resumes. The last three notices each moved the date later, which is not a pattern to rely on in reverse.
“A paused deprecation is a deadline with the date removed. The work it was asking for is the same size as before.”
Migrating changes behaviour, not just the token
The mechanical part of migration is small. For an app used in one workspace, Slack's migration guide describes a scope-selection screen in the app settings and a reinstall, with no code involved. That description is accurate and incomplete, because several things the classic bot did without asking now need an explicit scope, and one response changes shape. Slack lists them in a separate differences page.
| Behaviour | Classic app | Granular Slack app |
|---|---|---|
| Posting to a public channel it has not joined | Allowed | Needs chat:write.public |
| Overriding username or icon per message | Allowed | Needs chat:write.customize |
| Posting as bot or as user | chat:write:bot, chat:write:user, as_user | One chat:write scope; the token type decides |
| OAuth endpoints | oauth/authorize and oauth.access | oauth/v2/authorize and oauth.v2.access |
| Top-level token in the OAuth response | User token | Bot token; user token moves under authed_user |
| Slash commands and incoming webhooks | Tied to the installing user | Tied to the bot user |
| RTM API | Available | Not available |
The posting row is the one that shows up in production. An alerting bot that has written to #incidents for six years without being a member will, after migration, get an error on its first post unless you added chat:write.public or invited it to the channel. Slack's API names the missing scope in the error, so the fix takes minutes. Whether the on-call engineer sees that error before an alert goes missing depends on whether anything watches the bot's own failures.
The OAuth row is the one that corrupts data quietly. Under v1, the top-level access_token in the response was the user token. Under v2 it is the bot token, and the user token sits inside an authed_user object. Code that reads the top-level field and stores it as "the user's token" will keep working and will store the wrong credential.
The scope screen has its own trap. If your app used the umbrella bot scope, the migration tool pre-ticks every scope that the umbrella contained. Accepting the defaults reproduces the over-broad permissions under new names. Slack's guide says you probably do not need all of them and singles out the remote_files scopes as rarely required. It also says Slack cannot work out which scopes your app uses, so the audit is yours: list each method the bot calls and each event it listens for, and untick the rest.
One change is in your favour. Slash commands and incoming webhooks attach to the bot user after migration, so they no longer disappear when the employee who installed the app leaves and their account is deactivated. If you have ever lost a webhook to an offboarding, this is the reason to migrate that needs no deadline.
RTM is the part with no like-for-like replacement
Most classic apps that are still alive are alive because of the Real Time Messaging API. Granular apps cannot call it, and Slack's FAQ is plain that once classic apps go there will be no way to reach RTM at all. The suggested replacement for an internal bot behind a firewall is Socket Mode, which keeps the outbound WebSocket pattern: call apps.connections.open with an app-level token, connect to the URL it returns, receive Events API payloads.
Three RTM capabilities do not cross over. Slack's migration FAQ says the Events API does not expose the typing indicator, and that the team plan information returned by rtm.start has no equivalent. The 2024 announcement that blocked new classic apps adds presence subscriptions to the list of RTM quirks without a modern counterpart. A bot that reacts to "user is typing" or tracks who is online has a feature to redesign, not a transport to swap.
Order of operations for an app installed in more than one workspace
For a distributed app, the configuration change and the code change have to land in a particular order, and the guide's FAQ contains the constraints.
- Build a separate staging app first. It will be granular by necessity, since classic apps can no longer be created, and it lets you test scopes without touching the live configuration.
- Write and deploy the v2 OAuth handling before migrating the live app's configuration. Once the configuration is migrated, the v1 OAuth endpoint stops working for that app.
- Expect two token populations for a while. There is no automatic conversion of existing tokens outside the development workspace. Each workspace gets granular tokens only when someone there reinstalls, and classic bot tokens keep working alongside in the meantime.
- Decide whether to nudge or force reinstalls. Slack suggests a direct message with the new authorise link, and notes that auth.revoke will force the issue if you are comfortable with that.
Slack does offer a rollback to the previous scopes if you confirm the scope list with its team. It calls the process complicated and warns of delay, and says the commonest reason people need it is migrating the configuration before the development work was done. Treat the configuration step as one-way.
Set the date Slack removed
Admins can see every affected app under the Deprecations tab in workspace or organisation settings. Developers can find theirs on the app management page, labelled "Classic". Run that inventory this week, while the absence of a deadline means nobody is competing for the same reviewers.
Then give each app an owner and a date of your own choosing. The apps without an owner are the finding. Every classic app is more than two years old, many were installed by people who have since left, and a bot nobody can explain is a credential with broad scopes that nobody is watching. Slack may publish a new cutoff next quarter or never. The inventory is worth having in both cases, and it is the only part of this timeline you control.
Frequently asked questions
Related reading
Auth0 Rules and Hooks Are Read-Only. Which End-of-Life Date Applies Depends on Whose Page You Read.
Some trackers list 18 November 2026 as the day Auth0 Rules and Hooks stop. Auth0's docs give 30 April 2027 for Hooks and 31 January 2028 for Rules. Here is what to check and how to migrate.
Voice Dictation Runs Four Times Faster Than Typing. Adoption Still Lags.
Voice dictation tools now claim 95%+ accuracy at sub-second latency, and output text far faster than typing. Most knowledge workers still don't use them daily. The holdup is correction cost, not the model.
AI coding tools moved to metered pricing in 2026. Most engineering budgets didn’t move with them.
Three major AI coding tools switched to metered billing in 2026, and one vendor’s 24-hour near-miss on plan access showed how thin the flat fee always was. Here’s what actually drives the bill, and how to budget it.