Menu

Mattering Automations Upstash integration

Connect Upstash to Mattering Automations: create and configure Redis databases, take and restore backups, manage Vector indexes, run teams and membership, and read the audit log as workflow steps, authenticated with your account email and an Upstash management API key.

The Mattering Automations Upstash integration puts the Upstash Developer API on your automation canvas: create, configure, rename and delete Redis databases, set budgets, read regions, eviction, TLS and auto upgrade, take, list, restore and delete backups, create and manage Vector indexes, run teams and membership, and read the account audit log. All 37 endpoints the Developer API documents are built as 37 nodes.

The credential is your Upstash account email plus a management API key, sent as HTTP Basic where the email is the username and the key is the password. Mattering ships a form labelled Account Email and Management API Key rather than a generic username and password pair, because a field labelled Password invites the console password, which is not what this API wants and produces a 401 that says nothing about which half is wrong.

One boundary decides whether this integration is the right tool, so it is stated first. Upstash runs four HTTP APIs and no two of them share a credential: the Developer API, the Redis REST data API on each database's own host, the Vector REST data API on each index's own host, and QStash. This integration builds the Developer API only. It manages Redis databases and Vector indexes; it does not read or write Redis keys, does not upsert or query vectors, and does not publish QStash messages. The Developer API has no QStash endpoints and no Kafka endpoints at all, so nothing here manages those either.

Before you start

  • A NATIVE Upstash account. Upstash documents the Developer API as available only to native Upstash accounts and states that accounts created through third-party platforms such as Vercel or Fly.io are not supported. This is the single most important prerequisite on the page, because it is invisible from inside the product and there is no workaround.
  • Permission to create a management API key on that account, in the Upstash console under Account and then Management API.
  • An understanding that a management key is a full-account key. Upstash lists role-based access for keys as a roadmap item rather than an available feature, so a key created today can delete every database on the account. There is no way to scope it down.
  • A check of your plan against the node families you intend to use. Upstash gates the API by account ORIGIN rather than by plan, and both managed products document a free plan value, but the reference does not say which plan gates the security add-ons it enumerates and does not state whether backups and teams are available on every tier. Confirm on Upstash's pricing page before building a workflow that depends on the backup or team nodes.
  • A Mattering workspace with access to the Automations app.
  • Not required and not stored here: a database REST token, a Vector index token or a QStash token. Those are data-plane credentials for APIs this integration does not call.

Create an Upstash management API key

Before anything else, check the account itself. Upstash documents the Developer API as available only to native Upstash accounts and states that accounts created through third-party platforms such as Vercel or Fly.io are not supported. Someone in that position has a working console, working databases and working data-plane tokens, and simply no Management API page to create a key on, with no error anywhere that explains why. The only fix is a native Upstash account. With that settled, the key itself takes a minute.

If this Upstash account was created through Vercel or Fly.io, stop here. Upstash states that the Developer API is not available to accounts created via third-party platforms, so no key can be created and no amount of retrying will change it. Create a native Upstash account instead.
Sign in to the Upstash console and open Account.
Open Management API and create a key, giving it a distinct name such as Mattering Automations.
Copy the key now. Upstash states that it does not keep your API key, so a forgotten key is useless and has to be replaced with a new one. Note also that a management key carries full account access: role-based access for keys is on Upstash's roadmap rather than available, so this key can delete every database on the account.
Note the email address the Upstash account is registered with. It is the Basic username on every request, and it is not an account handle.

Connect Upstash to Mattering

In Mattering Automations, open Connectors and click Add Connector.
Choose Upstash from the list.
Click Create new credential, put the account email into Account Email and the management key into Management API Key, and save.
The credential is stored encrypted in Google Secret Manager and never appears in workflow logs. The two fields are labelled for what they are because both halves are easy to fill in wrongly: the username is an email address rather than a handle, and the password is a management key rather than your console password.
Click Test Connection. Mattering lists your Redis databases, which is the cheapest authenticated read on the API and takes no parameters at all, so it succeeds even on an account with nothing created yet. A wrong key returns 401.
Click Connect. All 37 Upstash nodes appear in your canvas palette.

Verify it works

Open any automation canvas and check the palette for an Upstash section. Drop Upstash: List Redis Databases onto it and run a test step. It takes no configuration, and it returns every Redis database on the account with its id, name, region, plan type, state and endpoint host. Note the response shape, because it is unusual and it decides how your templates read: Upstash list endpoints return a BARE JSON ARRAY with no wrapper key, so paths read data[].database_id rather than data.databases[].database_id.

Now check the other managed product. Drop Upstash: List Vector Indexes onto the canvas and run it: same bare-array shape, with each index's id, name, region, similarity function, dimension count and index type. One caution while you are looking at the output: this response also carries each index's live data-plane token and read-only token, with no option to hide them, so treat the run record as secret material.

Finish with a read that proves the credential reaches account-level data rather than just resource lists. Drop Upstash: List Audit Logs onto the canvas and run it: it returns who did what, when, from which IP and through which client, and distinguishes a console action from an API-key one. Everything in this path is read-only, so it cannot reach any of the seven destructive nodes and is safe to re-run whenever you replace the key.

Node reference

37 nodes, generated from the connector itself so this reference always matches what ships.

Read and query nodes 9

Upstash: Get Redis Database

upstash_get_redis_database

Get one Redis database's full configuration: name, region, state, endpoint host, port, plan type, budget, read regions, eviction and TLS settings, and its current limits. By default the response also carries the database's LIVE connection credentials, so treat this node's output as secret material or set the credentials field to hide.

Upstash: Get Redis Database Stats

upstash_get_redis_database_stats

Read usage and performance metrics for one Redis database: daily read and write request counts, connection count, keyspace size, throughput, disk usage, mean and 99th percentile latency, monthly bandwidth, monthly request count and month-to-date billing. The time-series fields are arrays of {x, y} points where x is a timestamp string. The window is fixed by Upstash and takes no parameters, so this node declares no incremental window.

Upstash: Get Team Members

upstash_get_team_members

List the members of one team, each with their email address and role (owner, admin, dev or finance). The response is a bare JSON array, so paths read data[].member_email.

Upstash: Get Vector Index

upstash_get_vector_index

Get one Vector index's configuration: name, region, similarity function, dimension count, embedding model, index type, plan and limits, plus its endpoint host. The response carries the index's live data-plane token, so treat the output as secret material.

Upstash: List Audit Logs

upstash_list_audit_logs

Read the account's audit trail: who did what, when, from which IP, and through which client. Each entry carries the actor's email, a human-readable action string, the affected entity, and the source, which distinguishes a console action from an API-key one. The response is a bare JSON array, so paths read data[].action_string. This is the node to build a change-alerting workflow on. Upstash documents no parameters for it, so there is no filtering or paging here; use extra_query if the vendor adds some.

Upstash: List Redis Backups

upstash_list_redis_backups

List the backups taken from one Redis database, each with its backup id, name, creation time, state and size. The response is a bare JSON array, so paths read data[].backup_id. This is what the backup picker on the restore and delete nodes reads from.

Upstash: List Redis Databases

upstash_list_redis_databases

List every Upstash Redis database on the account, with each one's id, name, region, plan type, state, endpoint host and current limits. The response is a bare JSON array, so paths read data[].database_id. This is the node every other Redis node's database picker reads from, and it is also the cheapest way to confirm a credential works.

Upstash: List Teams

upstash_list_teams

List every team the account belongs to, with each team's id and name. The response is a bare JSON array, so paths read data[].team_id. This is what the team picker on the membership, move-to-team and index-transfer nodes reads from.

Upstash: List Vector Indexes

upstash_list_vector_indexes

List every Upstash Vector index on the account, with each one's id, name, region, similarity function, dimension count, embedding models, index type, plan and limits. The response is a bare JSON array, so paths read data[].id. It also carries each index's LIVE data-plane token and read-only token, so treat this node's output as secret material. This is what the index picker on the other Vector nodes reads from.

Action nodes 28

Upstash: Add Team Member

upstash_add_team_member

Add a person to a team by email address and give them a role. They gain access to everything the team owns as soon as this returns, so treat the role as an access-control decision rather than a label.

Upstash: Change Redis Plan

upstash_change_redis_plan

Move a Redis database onto a different plan. This changes what the account is billed, immediately. It is reversible in the sense that the plan can be changed back, but money already spent is not returned by changing it back. Moving DOWN to a smaller fixed plan can fail if the database already holds more data than the target allows. Returns the string OK.

Upstash: Create Redis Backup

upstash_create_redis_backup

Take a point-in-time backup of a Redis database. This is the only thing that makes Upstash: Delete Redis Database recoverable, so run it before any bulk rewrite or teardown. The call returns the string OK and no backup id: read Upstash: List Redis Backups afterwards to find the new backup and watch its state reach completed.

Upstash: Create Redis Database

upstash_create_redis_database

Create a new Upstash Redis database. The response carries the new database's id and its endpoint host, and by default its live connection credentials as well, so treat the output as secret material. Choosing a paid plan starts billing immediately.

Upstash: Create Team

upstash_create_team

Create a team. Teams own databases and indexes separately from the personal account, which is how access and billing are shared. The response carries the new team id.

Upstash: Create Vector Index

upstash_create_vector_index

Create an Upstash Vector index. The similarity function and dimension count are fixed at creation and decide what embeddings the index can ever hold, so they must match whatever produces your vectors. The response carries the new index's endpoint host and its live data-plane token, which is what the Vector REST API needs - this connector manages indexes and does not read or write vectors itself.

Upstash: Delete Redis Backup

upstash_delete_redis_backup

DESTRUCTIVE AND IRREVERSIBLE. Delete one backup. The database and its live data are untouched; what is lost is a restore point, which matters most in exactly the situation a backup exists for. Other backups of the same database survive. Returns the string OK.

Upstash: Delete Redis Database

upstash_delete_redis_database

DESTRUCTIVE AND IRREVERSIBLE. Delete a Redis database and every key stored in it. The endpoint host stops resolving and every client using it fails. There is no trash and no undo: the only recovery is a backup taken beforehand with Upstash: Create Redis Backup, and a backup is destroyed along with its database. Backups of OTHER databases and the rest of the account are unaffected. Returns the string OK.

Upstash: Delete Team

upstash_delete_team

DESTRUCTIVE AND IRREVERSIBLE. Delete a team. Every member loses access through it and the team's billing arrangement goes with it. Move any database or index the team owns out first - Upstash: Move Redis Database To Team and Upstash: Transfer Vector Index move resources between owners. Returns the string OK.

Upstash: Delete Vector Index

upstash_delete_vector_index

DESTRUCTIVE AND IRREVERSIBLE. Delete a Vector index and every vector and piece of metadata in it. The endpoint host stops resolving. Upstash Vector has no backup endpoint in the developer API, so unlike a Redis database there is no restore point to fall back on: the only recovery is re-embedding and re-upserting the source data. Other indexes are unaffected. Returns the string OK.

Upstash: Disable Redis Auto Upgrade

upstash_disable_redis_auto_upgrade

Stop Upstash moving this database to a larger plan on its own. Spend stops growing unattended; in exchange, the database starts rejecting writes once it reaches the limits of its current plan. Any upgrade already applied stays applied. Returns the string OK.

Upstash: Disable Redis Daily Backup

upstash_disable_redis_daily_backup

Turn off automatic daily backups for a Redis database. No further automatic backups are taken; backups already on the account are not deleted by this call. Returns the string OK.

Upstash: Disable Redis Eviction

upstash_disable_redis_eviction

Stop the database evicting keys when it fills up. Nothing is restored, and once the size limit is reached new writes are rejected rather than making room. Returns the string OK.

Upstash: Enable Redis Auto Upgrade

upstash_enable_redis_auto_upgrade

Let Upstash move this database to a larger plan on its own when it outgrows the current one, rather than rejecting writes. This is a BILLING setting: an upgrade it performs raises the bill without asking. Reverse it with Upstash: Disable Redis Auto Upgrade. Returns the string OK. (The vendor's reference files this page under the name autoscaling; the endpoint is autoupgrade and this is what it does.)

Upstash: Enable Redis Daily Backup

upstash_enable_redis_daily_backup

Turn on automatic daily backups for a Redis database. Upstash then takes and retains them on its own schedule; retention is a console setting rather than a parameter here. Returns the string OK.

Upstash: Enable Redis Eviction

upstash_enable_redis_eviction

Let the database evict old keys when it reaches its size limit, instead of rejecting new writes. The setting itself is reversible, but keys evicted while it is on are gone for good, so this is a cache setting and not a storage one. Returns the string OK.

Upstash: Enable Redis TLS

upstash_enable_redis_tls

Require TLS on connections to this Redis database. ONE WAY: the developer API documents no disable-TLS endpoint, so this cannot be undone here. Any client still connecting without TLS will need reconfiguring. Returns the updated database object.

Upstash: Move Redis Database To Team

upstash_move_redis_to_team

Move a Redis database from the personal account into a team, transferring ownership and billing with it. The database keeps its id, endpoint host and credentials, so nothing connected to it breaks. Everyone in the target team gains access according to their role. Returns the string OK.

Upstash: Remove Team Member

upstash_delete_team_member

Remove a person from a team. Their access to everything the team owns ends immediately. Nothing is deleted and the person can be added back with Upstash: Add Team Member, which is why this is a reversible write rather than a destructive one - but note that the request body carries the team and the email only, so removing the wrong person is a single wrong picker selection away. Returns the string OK.

Upstash: Rename Redis Database

upstash_rename_redis_database

Change a Redis database's display name. This is a label only: the database id, the endpoint host and the connection credentials are all unchanged, so nothing connected to it breaks. Returns the updated database object.

Upstash: Rename Vector Index

upstash_rename_vector_index

Change a Vector index's display name. A label only: the index id, the endpoint host and the tokens are unchanged, so nothing connected to it breaks. Returns the string OK.

Upstash: Reset Redis Password

upstash_reset_redis_password

DESTRUCTIVE TO EVERY LIVE CLIENT. Rotate the database's password and REST token. No data is deleted and no key is touched, but every application, worker, cron job and deployed environment variable holding the previous credential stops authenticating the moment this returns, and the old value cannot be restored. Run it only when you are ready to redeploy the new credential everywhere. The new credentials come back in the response, which is therefore secret material.

Upstash: Reset Vector Index Tokens

upstash_reset_vector_index_passwords

DESTRUCTIVE TO EVERY LIVE CLIENT. Rotate the index's data-plane token AND its read-only token together. No vector is deleted and the index keeps its id, name and endpoint host, but every application holding either old token stops authenticating the moment this returns, and the old values cannot be recovered. Read the new ones back with Upstash: Get Vector Index. Returns the string OK, not the new tokens.

Upstash: Restore Redis Backup

upstash_restore_redis_backup

DESTRUCTIVE TO THE TARGET DATABASE. Restore a backup into a database. Upstash's own documentation states that all existing data in the target database is deleted before the restore begins, so anything written since the backup was taken is lost unless a second backup was taken first. The backup itself is not consumed and stays available, and other databases are untouched. The database id, endpoint host and credentials do not change, so clients stay connected - they simply see different data. Returns the string OK.

Upstash: Set Vector Index Plan

upstash_set_vector_index_plan

Move a Vector index onto a different plan. This changes what the account is billed, immediately. Reversible in that the plan can be changed back; money already spent is not returned by changing it back. Returns the string OK.

Upstash: Transfer Vector Index

upstash_transfer_vector_index

Move a Vector index to a team, transferring ownership and billing with it. The index keeps its id, endpoint host and tokens, so nothing connected to it breaks, and everyone in the target team gains access. ONE WAY in practice: Upstash states that transferring to a personal account is not supported, so an index moved into a team cannot be moved back out through this API. Returns the string OK.

Upstash: Update Redis Budget

upstash_update_redis_budget

Set the monthly spend ceiling for one Redis database. This is the guard rail on usage-based charges, so setting it too low can stop a database serving once it is reached. Returns the string OK.

Upstash: Update Redis Read Regions

upstash_update_redis_read_regions

Replace the set of read regions for a global Redis database. The list is a REPLACEMENT, not an addition: any region you leave out is removed, and its read replica goes with it. The primary region is set at creation and is not changed here. Adding regions adds cost. Returns the string OK.

Using the integration

On the automation canvas

Upstash nodes chain with every other node. A task moving to approved can provision a database with Upstash: Create Redis Database and record its endpoint host in a Mattering database, a nightly schedule can read Upstash: Get Redis Database Stats and open a task when month-to-date billing turns, and Upstash: List Audit Logs can feed a change-alerting workflow. The id fields are pickers wherever a lister exists: databases from Upstash: List Redis Databases, grouped by whether they are free or paid and described by region, indexes from Upstash: List Vector Indexes, teams from Upstash: List Teams, backups from Upstash: List Redis Backups scoped to the database chosen on the same form, and team members from Upstash: Get Team Members scoped to the team.

From AI assistants over MCP

Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and Matter, the assistant built into the platform, can drive the same surface. Upstash publishes MCP servers of its own and it is worth knowing which job each does. Upstash ships a Redis MCP server aimed at the data plane, whose tools run Redis commands against a database and search the Upstash documentation, and a broader Upstash MCP for account-level work including creating databases and reading QStash and Workflow logs. For an agent that runs Redis commands against your keys, that is the vendor's tool and this integration cannot do it, because it never touches a database host. What this integration adds is the management API in full, including the parts outside their server's scope, such as Vector index lifecycle, backups and restores, team membership, budgets and read regions, and the audit log, running as workflow steps on a schedule or a trigger with nobody present, on a credential the workspace holds rather than one developer's client config. Keep an agent on the listers, the reads and the create and configure nodes, and gate the seven destructive ones.

The management plane, and what it is not

Upstash runs four HTTP APIs and no two share a credential: the Developer API on api.upstash.com, the Redis REST data API on each database's own host with that one database's token, the Vector REST data API on each index's own host with that index's token, and QStash with its own token. This integration builds the first and only the first, so it creates and configures Redis databases and Vector indexes without reading or writing what is inside them. The alternative was a credential form carrying five values for what a user experiences as one account, four of which are invalidated whenever someone resets a password, and it would still have been a half span, because the management API has no QStash surface to reach. The three data APIs are ordinary bearer-token HTTP APIs, so a generic HTTP node reaches them today, and each can become its own connector later without changing anything here.

Backups, and the response shape that shapes the workflow

Upstash: Create Redis Backup is what makes a delete or a bulk rewrite survivable, and its response decides how the workflow is built: it returns the string OK with no backup id, so the only way to find the backup just created is Upstash: List Redis Backups, where the state field tells you whether it finished. Wait for that state before treating the backup as a safety net. Upstash: Enable Redis Daily Backup and Upstash: Disable Redis Daily Backup control the automatic schedule, with retention configured in the console rather than through this API. Upstash: Delete Redis Backup is flagged destructive because a restore point is precisely what is missed when it is gone, and note the sharper fact underneath: a backup does not survive its database, so deleting the database takes its backups with it.

Restoring is a destructive write to the target

Upstash: Restore Redis Backup is the node most likely to be misread as additive. Upstash's own backup documentation states that all existing data in the target database is deleted before the restore operation begins, so everything written since the backup was taken is lost unless a second backup was taken first. Take that second backup as the first step of any restore workflow. What is not affected is worth knowing too: the backup itself is not consumed and stays available, and the database keeps its id, endpoint host and credentials, so clients stay connected and simply start seeing different data, which is the quiet part of the change.

Rotating credentials, and what that costs

Upstash: Reset Redis Password rotates a database's password and REST token, and Upstash: Reset Vector Index Tokens rotates an index's token and its read-only token together, so a read-only consumer breaks alongside the writer. Neither deletes a single byte. What they do is break every live client the instant they return: every application, worker, cron job and deployed environment variable holding the previous value stops authenticating, with no grace period and no way back to the old value. Run them as part of a workflow that also distributes the replacements. The Redis node returns the new credentials in its response; the Vector one returns only the string OK, so read the new tokens back with Upstash: Get Vector Index.

Secret material in node output

Three read nodes put live credentials into workflow output, and they say so in their descriptions. Upstash: Get Redis Database returns the database's connection credentials by default, which the credentials parameter can set to hide. Upstash: List Vector Indexes and Upstash: Get Vector Index return each index's data-plane token and read-only token, with no hide parameter available at all. Several writes are in the same position by nature: Upstash: Create Redis Database returns the new credentials and Upstash: Reset Redis Password returns the rotated ones. The integration follows the vendor's own defaults rather than deviating quietly, so the handling is yours to choose: set credentials to hide where the option exists, and pass a value straight to whatever consumes it rather than leaving it in workflow state.

Billing settings you can automate, and one naming trap

Upstash: Update Redis Budget sets a monthly spend ceiling per database, which is the guard rail on usage-based charges and is worth setting deliberately, since a ceiling set too low stops the database serving once it is reached. Upstash: Change Redis Plan and Upstash: Set Vector Index Plan move resources between plans and take effect immediately; moving down to a smaller fixed plan can fail if the database already holds more data than the target allows. Upstash: Enable Redis Auto Upgrade lets Upstash move a database to a larger plan on its own rather than rejecting writes, which raises the bill without asking, and Upstash: Disable Redis Auto Upgrade stops that at the cost of writes being rejected at the current plan's limits. The naming trap is the vendor's own: the reference files those two pages under a name that suggests autoscaling, and the endpoints and the behaviour are auto UPGRADE, a billing setting rather than a capacity one.

One-way changes worth planning around

Three nodes are effectively one-way even though nothing is destroyed, so they are flagged in their descriptions rather than in the destructive list. Upstash: Enable Redis TLS has no disable counterpart anywhere in the reference. Upstash: Transfer Vector Index moves an index into a team, and Upstash states that transferring to a personal account is not supported, so it cannot be moved back out through this API. Upstash: Update Redis Read Regions replaces the whole set rather than adding to it, so any region omitted from the list is removed and its replica goes with it, which is easy to get wrong from a template that sends only the region being added. One more region detail: Redis and Vector do not share a region list, so a code copied from one to the other is a plausible and confusing failure.

The seven destructive nodes, split by what they cost

Conflating these two groups is the mistake worth avoiding. Data is destroyed by Upstash: Delete Redis Database, which takes the database and every key in it with no trash and no undo, recoverable only from a backup taken beforehand, and a backup does not survive its database; by Upstash: Delete Vector Index, which is worse, because the Developer API ships no Vector backup endpoint at all, so the only recovery is re-embedding and re-upserting the source data; by Upstash: Delete Redis Backup, which removes one restore point; by Upstash: Restore Redis Backup, which deletes the target's existing data before restoring; and by Upstash: Delete Team, which ends everyone's access through it, so move any database or index the team owns out first with Upstash: Move Redis Database To Team or Upstash: Transfer Vector Index. No data at all is touched by Upstash: Reset Redis Password and Upstash: Reset Vector Index Tokens, which break every live client instead. Gate all seven behind a condition or an approval task.

Troubleshooting

There is no Management API page in my Upstash console
The account was almost certainly not created directly with Upstash. Upstash documents the Developer API as available only to native Upstash accounts and states that accounts created through third-party platforms such as Vercel or Fly.io are not supported. Everything else about such an account works, which is what makes this confusing: the console opens, databases run and data-plane tokens authenticate, and only this page is missing, with no message explaining it. There is no setting to enable and no workaround inside this integration. Create a native Upstash account and use its key.
Test Connection fails with 401
One of the two halves is wrong, and Upstash's 401 does not say which. Check the username half first, because it is the more common mistake: it is the EMAIL ADDRESS the Upstash account is registered with, not an account handle or a display name. Then check the password half: it is a management API key created under Account and then Management API, not your Upstash console password. If the key is right and still rejected, create a new one, since Upstash states it does not keep your key and a partially copied value cannot be checked against anything.
The credential form was saved but a node reports a missing field
Because authentication here is HTTP Basic, a credential missing the email half fails inside Mattering's credential resolution and never reaches the network at all, so the node reports a configuration error rather than an HTTP status. Reopen the credential and confirm both Account Email and Management API Key are filled. A credential created earlier as a generic username and password pair still resolves, so an older connection keeps working, but a new one should use the Upstash form so the labels match what the values actually are.
A template reading a list of databases resolves to nothing
Check for a wrapper key that is not there. Every Upstash list endpoint returns a BARE JSON ARRAY, so the paths are data[].database_id, data[].team_id, data[].backup_id and data[].member_email, with no data.databases[] level in between. This is unusual compared with most integrations and it is the most common cause of a path that looks right and resolves to nothing. Run the lister on its own and read the output shape before writing the template.
A write node returned OK and I cannot find what it created
That is the documented response rather than a truncation: nine of the mutations answer with the bare string OK, carrying no id, no timestamp and no statement of what changed. Upstash: Create Redis Backup is the sharp case, because it returns OK with no backup id, so the only way to find the backup just taken is Upstash: List Redis Backups and to watch its state reach completed. Design the workflow as write then read back, rather than expecting an id in the response. The create-database and rename nodes are the exceptions that do return an object.
The audit log node returns 404
This is the one known wire-shape ambiguity in the integration and it comes from the vendor's own documentation. Every other operation is served from the versioned base path, while the audit log page's own server block declares the path without the version prefix, and the integration follows that page for that endpoint. If it 404s, the vendor's page is out of step with the live API. Report it and use the node's extra query escape hatch or a generic HTTP node against the versioned path in the meantime.
A workflow is being rate limited
Upstash publishes no rate-limit documentation for the Developer API at all: no requests-per-second ceiling, no quota figure and no retry-after convention. So assume a limit exists, treat the ceiling as unknown, and do not design against a number. The practical fixes are behavioural: do not poll Upstash: List Redis Databases in a tight loop, put a delay between iterations of a loop that touches many resources, and let the pickers resolve from the cached options read rather than re-running listers as steps. Separately, Upstash documents a default ceiling on how many API keys one user may hold, each needing a distinct name, which is a different limit entirely.
Read regions disappeared after an update
Upstash: Update Redis Read Regions takes a REPLACEMENT set rather than an addition, so any region left out of the list is removed and its read replica goes with it. A template that sends only the region being added therefore removes everything else. Read the current set with Upstash: Get Redis Database first and write the edited full list back. While you are there, note that Redis and Vector do not share a region list, so a code copied from one product to the other will not be accepted.
Every client stopped connecting and no data is missing
A reset node ran. Upstash: Reset Redis Password and Upstash: Reset Vector Index Tokens rotate credentials without touching data, so the database or index is intact and every client holding a previous value is locked out with no grace period and no way back to the old one. The way forward is redeployment: read the new Redis credentials from the reset node's own response, or the new Vector tokens from Upstash: Get Vector Index, and push them everywhere. Build rotation as a single workflow that rotates and distributes in one run rather than as a standalone step.
A restore lost data written today
That is the documented behaviour rather than a fault. Upstash states that all existing data in the target database is deleted before a restore begins, so Upstash: Restore Redis Backup is a destructive write to the target and anything written since the chosen backup was taken is gone unless a second backup was taken first. Make Upstash: Create Redis Backup the first step of every restore workflow, and wait for its state to report completed in Upstash: List Redis Backups before restoring. The database keeps its id, endpoint host and credentials through a restore, which is why clients stay connected and simply see older data.

Upstash is a trademark of its respective owner. Mattering uses platform names to identify the services these integrations connect to; no endorsement or affiliation is implied.

Start Free Explore our Products