Menu
Mattering Automations connector

An Uploadcare MCP that knows the difference between uploaded and stored

Connect one Uploadcare public key and secret key and 41 nodes land on the canvas: fetch a file server-side from a URL, make it permanent before its 24-hour clock runs out, tag it, attach metadata, search it back, group it, convert a document or a video, run virus scanning, label detection, moderation or background removal, and subscribe a webhook so the next event starts a workflow of its own.

41 ready-made nodes One credential to connect Works with Claude, ChatGPT, and Matter

What the Uploadcare MCP connector does

The Uploadcare connector for Mattering Automations is 41 nodes on Uploadcare REST API version 0.7 plus the Upload API: 19 read and 22 act, of which four are flagged destructive. Files, groups, metadata, tags and search sit on the REST plane along with webhook subscriptions, document and video conversion, and the four Add-Ons (ClamAV virus scanning, AWS Rekognition label detection, AWS Rekognition moderation and remove.bg background removal). The Upload plane carries the from-URL upload, its status poll, an upload-side file read and group creation. One project key pair authenticates both planes, and Uploadcare's own signed scheme is what goes on the wire.

Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and any MCP-capable assistant can use them, alongside Matter, the AI assistant built into the platform. An agent can hand Uploadcare a URL with Uploadcare: Upload File from URL, wait for Uploadcare: Get Upload Status to turn the token into a real file id, make it permanent with Uploadcare: Store File, and tag it with Uploadcare: Update File Tags so a person can find it later, without anyone opening the Uploadcare dashboard.

One fact about Uploadcare governs almost every workflow anyone writes on it, and it is worth stating in the first minute rather than the second week. Every file arrives TEMPORARY. Uploadcare deletes it within 24 hours unless it is stored, or unless the project has Auto file storing switched on. Nothing has to fail for the file to disappear: an automation that uploads and never stores loses the file the next day, having received HTTP 200 on every call it made. Uploadcare: Get Project reports whether that setting is on, and Uploadcare: Store File and Uploadcare: Store Files (Bulk) are what turn a successful upload into a file that still exists tomorrow. Storing is additive and destroys nothing, so it is safe to run at any point inside the window.

The safety modeling is stated rather than implied, and one path name is genuinely misleading. Four nodes are flagged destructive: Uploadcare: Delete File, Uploadcare: Delete Files (Bulk), Uploadcare: Delete Group and Uploadcare: Delete Webhook. There is no restore endpoint anywhere in version 0.7, so a delete is permanent. The trap is that deleting a file looks like the inverse of storing one and is not: the endpoint sets a removal timestamp and clears the original file URL, and version 0.7 retired the older delete path in its favour. Two further details matter for anything doing takedowns. Deleted files stay LISTABLE, so a surviving metadata row is not evidence the bytes came back. And a delete does not invalidate the CDN cache, so a deleted file can keep being served from an edge for as long as that entry lives.

One connection, three surfaces

  • Visual automations. Drag Uploadcare nodes onto the canvas and wire them to triggers, conditions, and other apps.
  • AI agents, yours or ours. These nodes speak MCP, the open protocol, so Claude, ChatGPT, and any MCP-capable assistant can use them, and Matter, the agent built into the platform, uses them natively.
  • The whole platform. Combine with native tasks, projects, and other connectors, no middleware in between.

Every Uploadcare node, ready to drop on the canvas

41 nodes, generated from the connector itself so this list always matches what ships. Expand a category to browse.

Read and query nodes 19

Uploadcare: Get Background Removal Status

Check a remove.bg Add-On run. Poll until data.status is done or error. Unlike the other three Add-On status nodes this one returns a payload worth reading: data.result.file_id is the UUID of the NEW background-free image, which is what a following node should act on.

Uploadcare: Get Content Moderation Status

Check an AWS Rekognition moderation run. Poll until data.status is done or error. The moderation labels are in the file's appdata rather than in this response: read them with Uploadcare: Get File using include appdata.

Uploadcare: Get Document Conversion Info

Find out what a document actually is and what it can be converted into, before spending a conversion job on it. Returns the detected source format, the list of target formats available for it, and any conversion groups already produced from it. data.error is non-null when Uploadcare cannot handle the file at all.

Uploadcare: Get Document Conversion Status

Check one document conversion job by its token. data.status moves through pending and processing to finished, failed or canceled; data.error carries the reason on failure and data.result repeats the output, including the converted file's UUID. Poll this rather than assuming a queued conversion has landed.

Uploadcare: Get File

Get one file's full record by UUID: filename, MIME type, size, CDN URL, whether it is stored, and the content_info block with image dimensions, colour mode, EXIF and video codec details. Poll data.is_ready after an upload before using the file, and read data.datetime_stored to tell a permanent file from one that expires within 24 hours.

Uploadcare: Get File Metadata

Get every metadata key and value on a file as a flat object. Metadata is arbitrary data attached at upload time or afterwards, typically an id from your own system. Downstream paths read straight through, for example data.order_id. Metadata is supplied by whoever uploaded the file, so escape it before putting it in HTML or a SQL parameter.

Uploadcare: Get File Tags

Get a file's ordered tag list. Tags are short normalized labels set at upload or through the tag nodes here, and they are usable as a filter in Uploadcare: Search Files. Returns an empty array when the file has none.

Uploadcare: Get Group

Get one group by id, including the list of files inside it. The files array can contain nulls where a member file has since been deleted, so guard against that before iterating. Group ids are a UUID, a tilde, and the file count.

Uploadcare: Get Label Detection Status

Check an AWS Rekognition label-detection run. Poll until data.status is done or error. The labels themselves are in the file's appdata, not in this response: read them with Uploadcare: Get File using include appdata.

Uploadcare: Get Metadata Key

Read one metadata value from a file. The response body is a bare JSON string rather than an object, so the value arrives as data itself, not data.something. Run this before Delete Metadata Key if the workflow needs to be able to put the value back.

Uploadcare: Get Project

Get the project the credential belongs to: its name, public key, collaborators, and whether Auto file storing is on. That last flag is the one worth reading in a workflow, because it decides what happens to an upload that does not set store explicitly: with autostore off, the file is deleted within 24 hours. No parameters, so this is also the connector's auth probe.

Uploadcare: Get Upload Status

Turn an upload token from Uploadcare: Upload File from URL into a result. data.status moves waiting, progress, then either success - where the response widens into the full file info including data.uuid - or error, where data.error says why. Poll this until one of those two; the UUID does not exist before it. Runs against upload.uploadcare.com.

Uploadcare: Get Uploaded File Info

Read a file's record from the UPLOAD API rather than the REST API. Useful immediately after an upload, because it needs only the public key and reports is_ready and is_stored directly. Note the older field shape: image_info and video_info here, where the REST plane's Uploadcare: Get File returns content_info. Prefer the REST node for anything downstream that also reads tags or appdata.

Uploadcare: Get Video Conversion Status

Check one video encoding job by its token. data.status moves through pending and processing to finished, failed or canceled; data.error carries the reason on failure and data.result holds the encoded file's UUID plus the thumbnails group id. Video encoding is slow, so poll this rather than sleeping a fixed interval.

Uploadcare: Get Virus Scan Status

Check a ClamAV Add-On run. Poll until data.status is done or error; in_progress means keep waiting. The verdict itself is not in this response - once the status is done, read the file with Uploadcare: Get File using include appdata and look at appdata.uc_clamav_virus_scan.data.infected.

Uploadcare: List Files

List the project's files, newest or oldest first, with each file's UUID, original filename, MIME type, size, CDN URL, storage timestamps, metadata, tags and content details. This is the node every other Uploadcare node leans on: it is the source for every file picker in this connector and the cheapest way to confirm a credential works. The response also carries totals for stored, unstored and removed files. Paging is cursor-based: follow data.next rather than incrementing an offset.

Uploadcare: List Groups

List the project's file groups, each with its id, creation time, file count, CDN URL and API URL. A group bundles already-uploaded files so they can be addressed together, for example everything a user submitted in one form. This is the source for every group picker in this connector. Paging is cursor-based: follow data.next.

Uploadcare: List Webhooks

List the project's webhook subscriptions with their id, event, target URL, active flag, payload version and signing secret. Note the response is a BARE ARRAY, so paths are data[].id rather than data.results[].id. This is the source for the webhook pickers in this connector.

Uploadcare: Search Files

Search the project's files by filename, metadata, MIME type, UUID or tags, with optional filtering on size, upload date and whether the file is an image. It is a POST because the query is a structured object, but it reads and changes nothing. At least one condition is required. All top-level conditions are ANDed; the only OR is inside a single exact field, inside query, or in tags.any, and the only NOT is tags.none or is_image false. Newly uploaded files are indexed asynchronously and may not appear immediately.

Action nodes 22

Uploadcare: Convert Document

Queue document conversions, for example DOCX to PDF or PDF pages to JPEG. Each entry in paths is a source UUID followed by the /document/ operator and the target format. Conversion is asynchronous: the response returns a token per accepted path for Uploadcare: Get Document Conversion Status, plus a problems map naming any path it rejected. The source file is never modified; each conversion creates a new file.

Uploadcare: Convert Video

Queue video encoding jobs: change format to mp4, webm or ogg, adjust quality and size, cut a segment, and generate thumbnails. Each entry in paths is a source UUID followed by the /video/ operator and any options. Asynchronous: the response returns a token per accepted path for Uploadcare: Get Video Conversion Status, along with the thumbnails group id. The source file is never modified.

Uploadcare: Copy File to Custom Storage

Copy a file, or a transformed version of it, out to an S3 bucket connected to the project as a custom storage. Returns an s3:// URL. Source files may be stored or merely uploaded but must not be deleted, and must not exceed 5 GB. HTTP 201 means the copy was written; HTTP 200 means a file with that name ALREADY EXISTED at the destination and nothing was overwritten, which is the response to check when a scheduled copy appears to succeed but the bucket never changes.

Uploadcare: Copy File to Uploadcare Storage

Copy a file, or a TRANSFORMED version of it, into the project's own storage as a new file with a new UUID. This is how a CDN transformation becomes a stored asset: pass a CDN path carrying operators as the source and the rendered result is what gets saved. The source may be stored or merely uploaded but must not be deleted. Size ceilings are 100 MB when the source URL carries transformation operators and 5 GB when it does not. The immediate 201 can report size 0 and application/octet-stream while the copy finishes; read the new UUID back with Uploadcare: Get File for the settled values.

Uploadcare: Create Group

Bundle already-uploaded files into a group so they can be addressed with one id and one CDN URL. Groups are IMMUTABLE: there is no add-to-group or remove-from-group call anywhere in the API, so changing the contents means creating a new group with a new id. Every file must belong to the same project as the group. Grouping does not store the files, so temporary members still expire on their own 24-hour clock. Runs against upload.uploadcare.com with the public key.

Uploadcare: Create Webhook

Subscribe a URL to an Uploadcare event so the project posts a notification when it happens. The target URL must be unique for each project and event-type pair, so the same receiver can serve several events but not the same event twice. Set a signing secret and Uploadcare signs each payload with HMAC-SHA256 in the X-Uc-Signature header as v1=<hex digest>, which is what lets the receiver reject forged calls. Request body is form-encoded, not JSON.

Uploadcare: Delete File

PERMANENTLY delete one file. The path says storage and the verb reads like the inverse of storing, but it is not: v0.7 retired DELETE /files/{uuid}/ and this endpoint took its place. It sets datetime_removed, nulls original_file_url, and there is NO restore endpoint anywhere in the API. The file's metadata row survives and stays visible through List Files with removed true, which is not the same as being recoverable. Uploadcare also states that this does not invalidate the CDN cache, so a deleted file can keep being served from an edge to anyone holding its URL.

Uploadcare: Delete Files (Bulk)

PERMANENTLY delete up to 100 files in one call. Same irreversibility as Uploadcare: Delete File, multiplied: there is no restore endpoint, and the CDN cache is not invalidated. The response splits into result (the deleted records) and problems (UUIDs that could not be deleted). A partial failure still returns HTTP 200.

Uploadcare: Delete Group

Delete a group object. THE FILES IN THE GROUP ARE LEFT INTACT: this removes only the grouping, not any content. It is still destructive because there is no restore endpoint and groups are immutable, so the recovery path is creating a NEW group from the same files, which mints a NEW id. Every stored group id and every group CDN URL pointing at the old one stops resolving. Returns HTTP 204 with no body.

Uploadcare: Delete Metadata Key

Remove one metadata key from a file. Returns HTTP 204 with no body. This is classed as a reversible write rather than a destructive one on purpose: it touches no file bytes, Uploadcare keeps no metadata history, and the exact value can be read beforehand with Uploadcare: Get Metadata Key and put back with one Update Metadata Key call.

Uploadcare: Delete Webhook

Unsubscribe and delete a webhook. It is keyed by TARGET URL and nothing else, while a target URL is only required to be unique per project and EVENT TYPE, so one call can remove several subscriptions that share a receiver across different events. Returns HTTP 204 with no body, which is the same answer whether one subscription went or three did. There is no undo: the subscription and any auto-generated signing secret are gone, and event delivery stops silently. To pause deliveries reversibly, set is_active false with Uploadcare: Update Webhook instead.

Uploadcare: Execute Background Removal

Run the remove.bg Add-On to strip the background from an image. Execution is asynchronous: this returns a request id, and Uploadcare: Get Background Removal Status returns the NEW file's id once it is done. The source image is not modified or replaced; a separate file is produced.

Uploadcare: Execute Content Moderation

Run the AWS Rekognition moderation Add-On against an image to detect unsafe content. Execution is asynchronous: this returns a request id for Uploadcare: Get Content Moderation Status. Moderation labels land in the file's appdata under aws_rekognition_detect_moderation_labels, each with a confidence score and a parent category such as Violence. This node flags content; it never deletes anything.

Uploadcare: Execute Label Detection

Run the AWS Rekognition label-detection Add-On against an image to tag what is in it. Execution is asynchronous: this returns a request id for Uploadcare: Get Label Detection Status. The detected labels are written into the file's appdata under aws_rekognition_detect_labels, each with a confidence score and parent categories, and are read back with Get File using include appdata.

Uploadcare: Execute Virus Scan

Run the ClamAV Add-On against a file. Execution is asynchronous: this returns a request id, and Uploadcare: Get Virus Scan Status turns that into a verdict. The result lands in the file's appdata under uc_clamav_virus_scan with an infected flag and, when positive, infected_with. WARNING: setting params to {"purge_infected": true} makes this node PERMANENTLY DELETE the file when the scan finds an infection. The flag is off by default and is the only way a node not marked destructive can destroy a file here.

Uploadcare: Replace File Tags

Replace a file's ENTIRE tag set. Any tag not in the new list is removed, and an empty array clears every tag. It is still a reversible write: the response returns the exact deleted list, so the previous state is recoverable from this node's own output with one follow-up call. Sending the same set again, in any case or order, writes nothing and returns empty added and deleted lists. Use Uploadcare: Update File Tags instead when you only want to add or remove a few.

Uploadcare: Store File

Make one file permanent. Every Uploadcare file is temporary when it arrives and is deleted within 24 hours unless it is stored, so this is the call that turns a successful upload into a file that still exists tomorrow. It destroys nothing and can be run at any point inside the 24-hour window. Returns the file's full record with datetime_stored set. There is no matching un-store endpoint: the only way out of the stored state is Uploadcare: Delete File, which is permanent.

Uploadcare: Store Files (Bulk)

Make up to 100 files permanent in one call. The response splits into result, the file records that were stored, and problems, a map of the UUIDs that could not be: incomplete ids, ids from another project, or files that no longer exist. A partial failure still returns HTTP 200, so check data.problems rather than the status code.

Uploadcare: Update File Tags

Add and remove tags on a file in one atomic operation, leaving everything else alone. Deletions are applied first, then additions, so a tag can be removed and re-added in the same call. Tags already present in add are skipped and tags absent from the file in delete are ignored, both silently. The request fails with 400 if the result would exceed 50 tags. An empty body is valid and simply returns the current state.

Uploadcare: Update Metadata Key

Set one metadata value on a file, creating the key if it does not exist (HTTP 201) or overwriting it if it does (HTTP 200). The request body is a bare JSON string, which the runner builds from the value field. Overwriting keeps no history, so read the old value first if the workflow may need it back.

Uploadcare: Update Webhook

Change a webhook subscription's target URL, event, active flag or signing secret. Fields left blank are not sent, so the rest of the subscription is untouched. Flipping is_active to false is the reversible way to stop deliveries without deleting the subscription and losing its signing secret. Request body is form-encoded, not JSON.

Uploadcare: Upload File from URL

Hand Uploadcare a publicly reachable URL and it fetches the file server-side. This is the automation-native way to get bytes into a project: no file ever passes through the workflow. Returns a TOKEN, not a file id - feed it to Uploadcare: Get Upload Status to get the UUID once the fetch completes (the response can also come back as finished file info directly when duplicate checking matched an earlier upload). Remember that an uploaded file is TEMPORARY: leave store blank and the project's autostore setting decides, and if that is off the file is gone within 24 hours. This node runs against upload.uploadcare.com with the project's PUBLIC key, not the signed REST scheme.

Connecting Uploadcare takes one credential

Both keys live in the Uploadcare dashboard on the project's API keys page, and they belong to one project, which is why this connector asks for no account or project identifier anywhere. The public key identifies the project and is the Upload plane's entire authentication as a request parameter. The secret key is never sent: it is the HMAC key that signs each REST request and, where the project has Signed Uploads enabled, each upload. Uploadcare documents two authorization schemes and only one of them is for production. Mattering implements the signed Uploadcare scheme, in which the Authorization header carries the public key and a signature computed over the verb, the body digest, the content type, a fresh timestamp and the full request URI. The Uploadcare.Simple scheme, which puts the project's secret key in cleartext in every request, is documented by the vendor as development only and is deliberately not implemented and not offered as an option here. Mattering also pins the versioned Accept header to 0.7, which is the only version Uploadcare currently supports; this matters more than it sounds, because deprecated versions still ANSWER with HTTP 200 and a DIFFERENT response shape rather than failing. One plan note before you build around the whole catalog: the four Add-Ons are third-party services billed through Uploadcare, and remote copy writes into a custom storage that has to be configured on the project first, so confirm what your plan includes on Uploadcare's pricing pages rather than assuming those eight Add-On nodes and the remote copy node will run.

1

Sign in to the Uploadcare dashboard, select the project this credential should reach, and open its API keys page. Copy both the public key and the secret key.

2

In Mattering Automations, open Connectors, click Add Connector, choose Uploadcare, and create a credential with both values.

3

Click Test Connection. Mattering signs a read of the project endpoint, which is the cheapest authenticated call in the API: no path parameters, no query string, no pagination. A pass proves the key pair is live, that it belongs to one project, and that the signature construction, the timestamp format and the pinned version header are all correct together.

4

Note what the pass does not cover. The Upload plane is a different host with different authentication and a different signature algorithm, so run Uploadcare: Upload File from URL once before trusting that half, particularly on a project with Signed Uploads enabled.

5

All 41 Uploadcare nodes appear in the canvas palette. Run Uploadcare: Get Project first and read whether Auto file storing is on, because that one flag decides whether your uploads survive their first day.

Common uses for the Uploadcare MCP

Ingest a file without ever touching its bytes

Uploadcare: Upload File from URL hands Uploadcare a publicly reachable address and Uploadcare fetches it server-side, which is the automation-native path: no file ever passes through the workflow. What comes back is a TOKEN rather than a file id, and the difference is the thing that trips people. Uploadcare: Get Upload Status turns that token into a result, and only when the status reaches success does the response widen into full file info carrying the real UUID. Then decide, in the same run, whether the file should survive: unless the project has Auto file storing on, Uploadcare: Store File is what keeps it past 24 hours.

Make files findable in a system that has no folders

Uploadcare has no directory tree, so tags and metadata are how anything gets found again. Uploadcare: Update File Tags adds and removes tags in one atomic operation and leaves everything else alone, which is the node to reach for by default, while Uploadcare: Replace File Tags swaps the whole set and is the one to be careful with. Uploadcare: Update Metadata Key attaches arbitrary key and value pairs, typically an id from your own system, so a file can be traced back to the record it belongs to. Uploadcare: Search Files then queries across filename, metadata, MIME type, UUID and tags with filtering on size, upload date and whether the file is an image. Three limits to design around: metadata is capped at 50 keys per file with 64-character keys and 512-character values, tags are normalized on write and capped at 50 per file, and search indexing is asynchronous, so uploading and immediately searching in the same run is a race.

Screen user uploads with the Add-Ons before anything is published

Four Add-Ons run as execute-and-poll pairs: Uploadcare: Execute Virus Scan with ClamAV, Uploadcare: Execute Label Detection and Uploadcare: Execute Content Moderation with AWS Rekognition, and Uploadcare: Execute Background Removal with remove.bg. Each execute call returns a request id rather than a result, and each has a matching status node. Read the results the right way: three of the four status nodes return only a status, and the actual labels or verdict are written into the file's appdata, which you read back with Uploadcare: Get File. The exception is Uploadcare: Get Background Removal Status, which returns the UUID of the NEW background-free image. One flag on the virus scanner deserves its own attention: it can be told to purge an infected file permanently, which makes it the only node outside the four destructive ones that can destroy a file. It defaults off.

Convert documents and video without a rendering pipeline of your own

Uploadcare: Convert Document queues conversions such as DOCX to PDF or PDF pages to JPEG, and Uploadcare: Convert Video queues encoding jobs that can change format, adjust quality and size, cut a segment and generate thumbnails. Both are asynchronous and both hand back a token for Uploadcare: Get Document Conversion Status or Uploadcare: Get Video Conversion Status. Uploadcare: Get Document Conversion Info is the read to run first, because it reports what a file actually is and which target formats are available for it, which is cheaper than spending a job to find out. Two response habits save a debugging session. Each entry in the paths list is a source UUID followed by the conversion operator, and a bare UUID with no operator lands in the response's problems map rather than its result. And a partly failed batch still answers HTTP 200, so branch on the problems map rather than on the status code.

Turn a CDN transformation into a stored asset

Uploadcare's image transformations are URL operators on the CDN rather than API endpoints, which means there is nothing to call for them. What there is, and what makes them usable from an automation, is Uploadcare: Copy File to Uploadcare Storage: hand it a path carrying the operators and it saves the rendered result as a NEW file with its own UUID. That is how a resize, a crop or a format change becomes a permanent asset with no bytes in the payload. Uploadcare: Copy File to Custom Storage does the same into an S3 bucket connected to the project, and carries one status trap worth knowing before you rely on it: 201 means copied, while 200 means a file with that name already existed and the destination was NOT overwritten, which is the failure mode where a scheduled copy reports success forever and the bucket never changes.

Let Uploadcare events start Mattering workflows

Uploadcare: Create Webhook subscribes a URL to a project event so Uploadcare posts a notification when it happens, and Uploadcare: List Webhooks and Uploadcare: Update Webhook manage what exists. Point the target at an inbound webhook trigger in Mattering and an Uploadcare event becomes the start of a workflow: a file arrives, the trigger fires, and the run tags it, screens it and files it away. Two facts shape how you manage them. A target URL is unique per project AND event type, so one receiver can legitimately serve several events. That is exactly why Uploadcare: Delete Webhook has a larger blast radius than it appears: it is keyed by target URL alone, so one call can unsubscribe several subscriptions sharing a receiver, and the response looks the same either way. The reversible alternative is switching a subscription inactive with Uploadcare: Update Webhook.

Example: Ingest, screen and shelve an inbound file so it is still there next week

Built on the Automations canvas in a few minutes, or described in plain language to Matter and built for you.

1

Trigger: a form submission or an inbound webhook arrives with a publicly reachable file URL.

2

Uploadcare: Upload File from URL fetches it server-side and returns a token. Uploadcare: Get Upload Status is polled until the status reads success, which is the point at which a real UUID exists to work with.

3

Uploadcare: Store File makes the file permanent. This step is the difference between a working automation and one that quietly loses everything it collected: without it, and without Auto file storing on the project, Uploadcare deletes the file within 24 hours while every call in the run returns 200.

4

Uploadcare: Execute Virus Scan runs, and Uploadcare: Get Virus Scan Status is polled until it reads done. The verdict is not in that response, so Uploadcare: Get File reads it back out of the file's appdata, and a condition stops the run and opens a task if anything was found.

5

Uploadcare: Execute Label Detection tags what is in the image, and once its status reads done, Uploadcare: Get File collects the labels from appdata and Uploadcare: Update File Tags writes the useful ones onto the file so search can find them later.

6

Uploadcare: Update Metadata Key attaches the id of the record this file belongs to in your own system, and the CDN URL plus the UUID land in a Mattering database row next to it.

Uploadcare MCP questions, answered

What Uploadcare operations are supported?

41 nodes on REST API version 0.7 plus the Upload API: 19 reads and 22 actions. Files: list, get, search, store, delete, bulk store, bulk delete, copy to Uploadcare storage and copy to custom storage. Metadata: read the whole map, read one key, set one key, delete one key. Tags: read, replace the set, and add and remove atomically. Groups: list, get, create and delete. Webhooks: list, create, update and delete. Project: read the project the credential belongs to. Add-Ons: execute and poll for virus scanning, label detection, content moderation and background removal. Conversion: document info, convert documents, convert video, and a status poll for each. Upload plane: upload from a URL, poll its status, read a file from the upload side, and create a group. Deliberately not included: direct multipart upload of raw bytes, which needs a file a canvas field cannot hold, and the Uploadcare.Simple authentication scheme, which the vendor itself does not recommend for production.

Does this work with Claude, ChatGPT, or other AI assistants?

Yes. Mattering exposes these nodes over MCP, an open protocol for connecting AI assistants to tools, so any MCP-capable assistant, including Claude and ChatGPT, can connect to your workspace and use them. Matter, the assistant built into the platform, uses the same nodes natively.

Uploadcare publishes its own MCP server. Why use this connector?

It does, and for some jobs a vendor server is the better answer. If what you want is an assistant helping you work with Uploadcare inside your editor or your own client config, use Uploadcare's server. This connector is shaped for operations instead: 41 nodes across both planes, including the parts nobody reaches for while coding and everybody needs in production, such as bulk storing, webhook subscriptions, conversion polls and the four Add-Ons. They run as steps on an automation canvas as well as over MCP, so a screening job can run on a schedule or off a trigger with nobody present, chain in the same workflow with every other connector in the catalog, hold the credential once in a shared workspace store rather than in each person's local config, and carry destructive flags you can gate on. Neither replaces the other, and nothing here disparages Uploadcare's own tooling.

Why do my uploaded files disappear?

Because storing is what stops a 24-hour clock. Uploadcare deletes a file within 24 hours of arrival unless it is stored or the project has Auto file storing on, which Uploadcare documents as ON by default, so confirm it with Uploadcare: Get Project rather than assuming either way. This is the single most common way an Uploadcare automation fails, and it fails silently: every call in the run returns a success. Run Uploadcare: Get Project once and read whether Auto file storing is enabled on your project. If it is not, put Uploadcare: Store File, or Uploadcare: Store Files (Bulk) for up to 100 at a time, into the same workflow that uploads. Storing is additive and destroys nothing, so there is no cost to running it early. One related trap: creating a group does not store its members, so temporary files keep their own clock and can expire out of a group.

Can an AI agent delete my files by accident?

The four destructive nodes are flagged so you can gate them: Uploadcare: Delete File, Uploadcare: Delete Files (Bulk), Uploadcare: Delete Group and Uploadcare: Delete Webhook. There is no restore endpoint anywhere in version 0.7, so none of them has an undo. Two of them behave differently from how they read. Deleting a group leaves the member FILES intact and removes only the grouping, but groups are immutable, so the only route back is a new group with a new id, and every stored group id and group CDN URL pointing at the old one stops resolving. Deleting a webhook is keyed by target URL alone while a target URL is unique only per project and event type, so one call can unsubscribe several subscriptions that share a receiver. The reversible alternative there is setting the subscription inactive with Uploadcare: Update Webhook. Keep agents on the read nodes, the store nodes and the tag nodes, and put the four deletes behind a condition or an approval task. One more node deserves the same treatment despite not being flagged: Uploadcare: Execute Virus Scan can be told to purge an infected file permanently, and while that flag defaults off, it is the only non-destructive node here that can destroy a file.

Does deleting a file take it off the CDN?

Not immediately, and Uploadcare says so on both delete endpoints: a delete does not invalidate the CDN cache. So a deleted file can keep being served from an edge for as long as that cache entry lives, which matters for anything doing a takedown. The metadata behaves differently again: deleted files stay LISTABLE, and the file list can be asked for removed files with a count of them alongside the stored and unstored totals. That surviving row is not the file coming back. The bytes and the original file URL are gone and there is no restore endpoint, so read a listable row as a record of what was deleted rather than as evidence anything is recoverable.

Why do the upload and conversion nodes hand back a token instead of a result?

Because everything asynchronous on Uploadcare returns a handle and not a result, and each has its own poll node. Uploadcare: Upload File from URL returns a token that is explicitly NOT a file id, and Uploadcare: Get Upload Status turns it into one once the status reads success. The Add-On execute nodes return a request id for their matching status node. Document and video conversions return tokens for Uploadcare: Get Document Conversion Status and Uploadcare: Get Video Conversion Status. The practical rule is that the UUID does not exist until the poll says it does, so nothing downstream should reference it earlier. A second habit belongs with the first: partial success answers HTTP 200. Bulk store, bulk delete and both conversion nodes split their response into a result and a problems map, so a workflow branching on the status code will treat a half-failed batch as a clean success.

Do I need a separate credential for each project?

One credential per Uploadcare project. The key pair is scoped to a project, so it already selects which files you are working with, which is why no node here asks for a project identifier and why Uploadcare: Get Project can simply report which project answered. If you keep production and staging in separate Uploadcare projects, connect a credential for each and choose the right one on the nodes. Within a project, the same pair covers both the REST plane and the Upload plane, even though they are different hosts with different authentication: the public key is the Upload plane's whole identity as a request parameter, while the secret key signs REST requests and, where Signed Uploads is enabled, uploads too.

Is there a file picker on the fields that take a UUID?

Yes, on most of them, and it is honest about its ceiling. Fields taking a file UUID are populated from Uploadcare: List Files, showing the original filename with the MIME type underneath, and group and webhook fields are populated the same way from their own listers. The limit is Uploadcare's rather than ours: the file list has no text filter at all, so the picker shows one page of newest-first results and cannot be searched within. On a project with a very large number of files, an older UUID has to be pasted or templated in from an upstream node, and every one of those fields stays typeable for exactly that reason. Uploadcare: Search Files does have real full-text search, but it refuses to run without at least one condition and its terms need four or more characters, so it cannot serve a picker that has to populate before you type anything.

41 nodes waiting on the canvas

Put Uploadcare to work inside your workspace

Connect once, then use every node in visual automations and AI agent workflows across the Mattering platform.

All product names, logos, and brands are property of their respective owners and are used for identification purposes only. Use of these names, logos, and brands does not imply endorsement or affiliation. Uploadcare is a product of its respective owner; Mattering is an independent platform.

Start Free Explore our Products