Menu
Mattering Automations connector

A Tumblr MCP that reads the whole network and writes to your own blogs

Connect one Tumblr credential and 37 nodes land on the canvas. A free consumer key on its own reads any public blog, its notes, its likes, its pages and the network-wide tag feed. Add the OAuth values and the same credential unlocks your dashboard, drafts, queue, submissions, notifications and blocks, plus publishing, editing, reblogging, liking, following, blocking and queue management. Tumblr's OAuth 1.0a tokens never expire, so this is a one-time setup with nothing to refresh.

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

What the Tumblr MCP connector does

The Tumblr connector for Mattering Automations is 37 nodes on Tumblr API v2, split 22 reads and 15 actions. Tumblr's own reference splits its surface the same way the connector does. Seven endpoints need nothing but an OAuth consumer key sent as an api_key parameter: blog info, a blog's public posts, a post's notes, a blog's public likes, custom pages, avatars, and the network-wide tag feed. The other thirty are user-scoped and signed: the dashboard, your own likes and following, drafts, the queue, submissions, notifications, blocks, and every write. Both halves come from one credential, and every response arrives inside Tumblr's own wrapper, so every output path on every node starts at data.response.

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 watch a tag across the whole of Tumblr with Tumblr: Get Tagged Posts, turn a post's engagement into a list of real blogs with Tumblr: Get Post Notes, and leave a drafted response in your blog's drafts with Tumblr: Create Text Post for a person to approve, without anybody opening the Tumblr dashboard.

The credential degrades on purpose rather than failing, which is unusual enough to be worth explaining. A saved credential holding only the consumer key is valid: the seven public read nodes run normally, and Test Connection reports a read-only connection and names the three values that are missing. The other thirty nodes refuse locally, before any request leaves Mattering, with an error naming the missing fields and where to get them, rather than firing a request that comes back as an opaque 401. Fill all four values and the runner signs every request, including the public ones, which is not redundant: an authenticated read of somebody else's blog carries your own state on it, such as whether you already follow or already liked.

Nothing about the credential has to be maintained after setup. Tumblr's OAuth 1.0a tokens never expire and never rotate, so there is no consent round-trip at run time, no refresh ladder, and no token to write back into the secret store. That is the whole reason this connector signs with OAuth 1.0a rather than OAuth 2.0: Tumblr's OAuth 2.0 issues a rotating refresh token, and a rotation that a cold start misses leaves a dead credential with no fallback.

The safety modeling is stated rather than implied. Nine nodes are flagged destructive, and the reason is the network rather than the API. Tumblr is public and federated: a published post reaches every follower's dashboard immediately and can be reblogged within seconds, and a reblog is an INDEPENDENT post on somebody else's blog carrying its own copy of your content. Deleting your post removes it from your blog and from the tag feeds, but it cannot recall a reblog, cannot un-send the notification it already caused, and cannot un-fetch a copy already taken. Three nodes deliberately are NOT flagged, and that is a judgement rather than an oversight: on Tumblr a like, a follow and a block are server-side state on the account, so Tumblr: Unlike Post, Tumblr: Unfollow Blog and Tumblr: Unblock Blog genuinely restore the previous state when re-done. The flag means this cannot be taken back, and here it can.

One connection, three surfaces

  • Visual automations. Drag Tumblr 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 Tumblr node, ready to drop on the canvas

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

Read and query nodes 22

Tumblr: Check Followed By

Ask whether one specific blog follows one of yours, answered as a single boolean at data.response.followed_by. Much cheaper than paging a large follower list to find out, which is what makes it the right node inside a loop over a list of prospects. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Account Following

List the blogs the connected account follows, at data.response.blogs[], with a total at data.response.total_blogs. Each row carries the blog's name, URL, description and last-updated timestamp, so this doubles as a way to spot followed blogs that have gone quiet. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Account Info

Report the account this connection is authenticated as, and list every blog it can act as at data.response.user.blogs[] with each blog's name, title, URL, follower count, queue depth and draft count. This is the node that powers the blog picker on every write node in this connector, and the one to run first when another node returns an unexpected 401. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Account Likes

Read the posts the connected account has liked, whether or not the account shares its likes publicly - which is what makes this different from Tumblr: Get Blog Likes. Use it to turn a reading habit into a curated digest, or to sync liked posts into a datatable. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Account Limits

Read how much of the account's daily write budget is left: posts, photos, follows, likes, new blogs and video minutes, each with its ceiling, its remaining count and when it resets. Run this BEFORE a loop that posts or follows, because Tumblr rejects the request that crosses the line rather than slowing down, and the workflow has no other way to see it coming. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Blog Avatar

Get the URL of a blog's avatar at a chosen size. Useful for enriching a contact record or building a digest with pictures. Tumblr answers this one two different ways - a redirect to the image for an unsigned request, JSON for a signed one - so this node normalizes both into data.response.avatar_url and never downloads the image itself.

Tumblr: Get Blog Followers

List the blogs following one of YOUR blogs, at data.response.users[], with a total at data.response.total_users. Each row says whether you follow them back. Only works on a blog the credential owns - a blog's follower list is private to its owner. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Blog Following

List the blogs one of YOUR blogs follows, at data.response.blogs[]. This is the per-blog view; Tumblr: Get Account Following is the account-wide one and returns the same shape for the primary blog. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Blog Info

Read any public blog's profile: title, description, post and like counts, avatar, ask settings, its UUID, and whether the connected account already follows it. Works on a consumer-key-only connection, so this is the node to run first when checking that a Tumblr credential is wired correctly. Tumblr wraps every response, so the payload is under data.response. The blog object is at data.response.blog.

Tumblr: Get Blog Likes

Read the posts a blog has publicly liked. Works on a consumer-key-only connection, but ONLY for blogs that have turned on 'share likes'; a blog with likes hidden returns a 403 no matter how the request is authenticated, which is a privacy setting rather than a credential problem. Offset cannot go past 1000, and before/after/offset are mutually exclusive.

Tumblr: Get Blog Posts

Read a blog's published posts, newest first. This is the workhorse read: watch a competitor's blog, mirror your own archive, or pull the posts behind a tag on one blog. Works on a consumer-key-only connection. Each post carries its id_string and reblog_key, which is the pair every write node needs, and note_count for engagement. Set reblog_info or notes_info to enrich the rows at the cost of a slower call.

Tumblr: Get Dashboard

Read the connected account's own dashboard: the posts from every blog it follows, newest first. The Tumblr equivalent of a home timeline, and the input to a daily digest or a keyword-watching workflow. Page with since_id to fetch only what is new since the last run, which is cheaper and more reliable than offset. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Draft Posts

Read one of your blogs' unpublished drafts. Pairs with Tumblr: Create Post writing with state=draft, so an AI workflow can propose posts for a human to review in the Tumblr app rather than publishing them itself. Drafts page by before_id, not by offset. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Filtered Tags

List the tags the connected account has filtered out of its dashboard and search results. Read it before adding another with Tumblr: Filter Tag, because Tumblr caps the list at 1,000 tags and rejects the write that would exceed it. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Notifications

Read the activity feed for one of your blogs: likes, reblogs, replies, follows, mentions, asks and answered asks, newest first. Narrow it with types to build a single-purpose monitor - a new-follower alert, or a mention watcher - instead of filtering a firehose downstream. Page with before and the server-supplied data.response._links.next. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Post

Read ONE post in Neue Post Format: a content[] array of typed blocks plus the layout that arranges them, which is the shape Tumblr: Edit Post expects back. Use this to read a post before editing it, so the edit rewrites the real content rather than replacing it with a guess. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail. For a public read that works without OAuth, use Tumblr: Get Blog Posts with an id instead.

Tumblr: Get Post Notes

List a post's notes: who liked it, who reblogged it and who replied, each with the blog name and a timestamp. This is how engagement becomes a list of PEOPLE rather than a number, and it works on a consumer-key-only connection. Page backwards with before_timestamp taken from the oldest note on the previous page.

Tumblr: Get Queued Posts

Read the posts waiting in one of your blogs' queues, in the order they will publish, each with its scheduled_publish_time. The node for a workflow that tops the queue up when it runs low, or that reports on what is about to go out. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Submissions

Read posts submitted to one of your blogs by other people and still awaiting moderation, including the submitter's name and email when they were not anonymous. The intake side of a community blog: route new submissions into a review queue, or alert on them. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Get Tagged Posts

Read posts from ACROSS Tumblr carrying one tag, newest first. This is the brand-monitoring and trend-watching node: it searches the whole network rather than one blog, and it works on a consumer-key-only connection. Page backwards with the before timestamp; there is no offset here, so a scheduled monitor should remember the newest timestamp it saw and stop when it reaches it again.

Tumblr: List Blocked Blogs

List the blogs one of your blogs has blocked, at data.response.blocked_tumblelogs[]. Use it to audit a block list, or to check before blocking again. This endpoint carries its own tighter rate limit of 60 requests a minute, separate from the connector-wide budget. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: List Blog Pages

List a blog's custom pages - the standalone About, FAQ or contact pages that sit outside the post stream - with each page's URL, title and body in one call. Works on a consumer-key-only connection. Useful for mirroring a blog's static content or checking that an About page still says what it should.

Action nodes 15

Tumblr: Block Blog

DESTRUCTIVE. Block another blog from one of your blogs. It immediately loses the ability to follow you, reblog you, message you or see your blog, and any existing follow is severed - unblocking does NOT restore the follow, so the relationship is gone even after Tumblr: Unblock Blog. Block by blog name, or by a post id to block whoever posted it. This endpoint carries its own limit of 60 requests a minute. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Create Post

DESTRUCTIVE. Publish any kind of post to one of your blogs using Neue Post Format: a content[] array of typed blocks (text, image, link, audio, video, quote) plus an optional layout describing how they are arranged. This is the full surface; Tumblr: Create Text Post is the shortcut for the common case. Tumblr is a public network. A published post appears on the dashboard of every follower immediately and can be reblogged within seconds, and a reblog is an INDEPENDENT post on somebody else's blog carrying its own copy of the content. Deleting your post later removes it from your blog and from the tag feeds, but it cannot recall a reblog, cannot un-send the notification it already caused, and cannot un-fetch copies already taken. Deleting is one-way: the post cannot be restored, only written again as a new post with a new id, so anything holding the old post id or permalink points at nothing. Counts against the account's daily write budget: 250 published posts, 250 image uploads, 200 follows and 1,000 likes per day. Every request also counts against the consumer key's 1,000 per hour and 5,000 per day. Loops that write per item should pace themselves; Tumblr: Get Account Limits reports what is left. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Create Text Post

DESTRUCTIVE. Publish a text post to one of your blogs from a title and a body, without assembling Neue Post Format blocks by hand: the title becomes a heading block and each blank-line-separated paragraph of the body becomes its own text block. Tumblr is a public network. A published post appears on the dashboard of every follower immediately and can be reblogged within seconds, and a reblog is an INDEPENDENT post on somebody else's blog carrying its own copy of the content. Deleting your post later removes it from your blog and from the tag feeds, but it cannot recall a reblog, cannot un-send the notification it already caused, and cannot un-fetch copies already taken. Deleting is one-way: the post cannot be restored, only written again as a new post with a new id, so anything holding the old post id or permalink points at nothing. Set state to draft or queue when a workflow should propose a post rather than publish one, which is the safe default for AI-written content. Counts against the account's daily write budget: 250 published posts, 250 image uploads, 200 follows and 1,000 likes per day. Every request also counts against the consumer key's 1,000 per hour and 5,000 per day. Loops that write per item should pace themselves; Tumblr: Get Account Limits reports what is left. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Delete Post

DESTRUCTIVE. Delete a post from one of your blogs. WHAT IT DOES: removes the post from your blog, from your archive and from the tag feeds. WHAT IT DOES NOT DO: recall the reblogs. Every reblog is an independent post on somebody else's blog holding its own copy of the content, and it survives your delete untouched, as do the notifications the post already sent. Deleting is one-way: the post cannot be restored, only written again as a new post with a new id, so anything holding the old post id or permalink points at nothing. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Edit Post

DESTRUCTIVE. Replace the content of an existing post on one of your blogs. The content array REPLACES what was there; it does not merge, so read the post with Tumblr: Get Post first and send back a modified copy rather than a partial one. Tumblr is a public network. A published post appears on the dashboard of every follower immediately and can be reblogged within seconds, and a reblog is an INDEPENDENT post on somebody else's blog carrying its own copy of the content. Deleting your post later removes it from your blog and from the tag feeds, but it cannot recall a reblog, cannot un-send the notification it already caused, and cannot un-fetch copies already taken. Editing does not reach the reblogs: every reblog of the post keeps the content it copied at the time, so a correction reaches your blog and not the copies. The previous version cannot be recovered through the API. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Filter Tag

Add a tag to the connected account's filtered list, so posts carrying it stop appearing on the dashboard and in search. Private to the account, nothing is published, and the filter can be removed again in the Tumblr app - a reversible write. Tumblr caps the list at 1,000 tags, so read Tumblr: Get Filtered Tags first in a loop. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Follow Blog

DESTRUCTIVE. Follow a blog as the connected account. This is a PUBLIC action: the blog is notified immediately, your account appears in its follower list, and its posts start arriving on your dashboard. Tumblr: Unfollow Blog reverses the state but cannot recall the notification, and re-following sends another one. Counts against the account's 200 follows a day, which a bulk-follow loop will exhaust quickly. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Like Post

DESTRUCTIVE. Like somebody else's post as the connected account. This is a PUBLIC action on another person's blog: your blog name is added to that post's notes for anyone to read, and the author is notified immediately. Tumblr: Unlike Post removes the note but cannot recall the notification. Counts against the account's 1,000 likes a day. Needs the post's id AND its reblog key, both of which every read node returns. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Mute Post Notifications

Stop notifications from one of your own posts. Nothing is published and nobody else can tell: it is a private setting on your account, which is why this is a reversible write rather than a destructive one. The right node for a post that has gone viral and is drowning the notification feed a workflow reads. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Reblog Post

DESTRUCTIVE. Reblog somebody else's post onto one of your blogs, optionally with your own commentary added underneath. Tumblr is a public network. A published post appears on the dashboard of every follower immediately and can be reblogged within seconds, and a reblog is an INDEPENDENT post on somebody else's blog carrying its own copy of the content. Deleting your post later removes it from your blog and from the tag feeds, but it cannot recall a reblog, cannot un-send the notification it already caused, and cannot un-fetch copies already taken. A reblog also notifies the original author and appears permanently in the original post's notes, and deleting your reblog later removes your copy but not that note or that notification. Needs the source post's id AND its reblog key, which every read node returns together. Counts against the account's daily write budget: 250 published posts, 250 image uploads, 200 follows and 1,000 likes per day. Every request also counts against the consumer key's 1,000 per hour and 5,000 per day. Loops that write per item should pace themselves; Tumblr: Get Account Limits reports what is left. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Reorder Queued Post

Move one queued post to a different position in a blog's queue. Nothing is published and nothing is visible to anyone until the queue reaches the post, and the move can be undone by moving it back, so this is a reversible write. Use it to promote a timely post ahead of an evergreen backlog. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Shuffle Queue

DESTRUCTIVE. Randomize the order of every post in a blog's queue. Nothing is published by this node, so the harm is not the public kind - it is that the previous order is GONE. There is no undo and no record of what the order was, so a hand-curated sequence cannot be recovered, only rebuilt post by post with Tumblr: Reorder Queued Post. Read Tumblr: Get Queued Posts first if the order matters. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Unblock Blog

Remove a block from one of your blogs, or clear every anonymous block at once. The block itself is lifted completely, which is why this is not flagged destructive - but note it does NOT restore a follow the block severed. Skipped by the endpoint test matrix because it acts on a real third party and there is no sandbox. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Unfollow Blog

Stop following a blog as the connected account, removing your account from its follower list and its posts from your dashboard. Reversible - following again fully restores the state - so it is not flagged destructive, but the original follow notification cannot be recalled and re-following sends a new one. Skipped by the endpoint test matrix because there is no Tumblr sandbox and the target is a real blog. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Tumblr: Unlike Post

Remove the connected account's like from a post, taking your blog name back out of that post's notes. The state itself is reversible - liking again fully restores it - which is why this is not flagged destructive, but the notification the original like sent cannot be recalled, and liking again sends a new one. Skipped by the endpoint test matrix because there is no Tumblr sandbox and the target is a real person's post. Needs the full OAuth 1.0a credential (consumer key, consumer secret, token and token secret). A consumer-key-only connection can read public Tumblr but cannot run this node, and the node says so rather than sending a request that would fail.

Connecting Tumblr takes one credential

Register an application at tumblr.com/oauth/apps, which is free, immediate and needs no review. That page shows the OAuth Consumer Key and the OAuth Consumer Secret. The consumer key alone is a valid credential and unlocks the seven public read nodes. For the other thirty, authorize your own account in Tumblr's developer API console at api.tumblr.com/console, which runs the three-legged flow for you and shows the resulting OAuth token and token secret, or run the standard three-legged OAuth 1.0a flow once yourself. Because Tumblr's tokens never expire, that is a one-time step with nothing to redo later. Two limits are worth reading before you design a loop. Tumblr publishes a ceiling of 1,000 requests an hour and 5,000 a day against the consumer key, shared across everything using that registered application, which is an order of magnitude below its per-IP ceiling of 300 a minute and is the one that actually binds. Tumblr also publishes a per-account daily write budget of 250 published posts, 250 image uploads, 200 follows and 1,000 likes, and the blocks endpoints carry their own ceiling of 60 requests a minute. Tumblr: Get Account Limits reads what is left of the write budget, so a loop can check before it starts.

1

Register an application at tumblr.com/oauth/apps and copy the OAuth Consumer Key. If public reads are all you need, stop here.

2

For the user-scoped half, copy the OAuth Consumer Secret from the same page, then authorize your own account in the API console at api.tumblr.com/console and copy the OAuth token and token secret it shows.

3

In Mattering Automations, open Connectors, click Add Connector, choose Tumblr, and create a credential with the values you have.

4

Click Test Connection. With all four values Mattering signs a request to Tumblr's account endpoint and reports the account name and the blogs it can act as, so a credential pointed at the wrong Tumblr account fails loudly instead of quietly returning somebody else's perfectly good data. With the consumer key alone it reads the tag feed instead and reports a read-only connection, naming the three values that are missing.

5

All 37 Tumblr nodes appear in the canvas palette. On the user-scoped nodes the blog field is a picker sourced from Tumblr: Get Account Info, so it offers exactly the blogs this credential can act as.

Common uses for the Tumblr MCP

Watch a tag across the whole network, not just one blog

Tumblr: Get Tagged Posts reads posts from across Tumblr carrying one tag, newest first, and it runs on a consumer-key-only connection, so brand monitoring costs nothing more than a registered application. There is no offset on that feed, so a scheduled monitor remembers the newest timestamp it saw and pages backwards with before until it reaches that mark again. Pair it with Tumblr: Get Blog Info to size the blog behind an interesting post, and with Tumblr: Get Post Notes to see who is already amplifying it.

Let a workflow propose posts and a person publish them

Tumblr: Create Text Post takes a title and a body and assembles Neue Post Format for you, and Tumblr: Create Post takes the full block array when a post needs images, links, audio or quotes arranged deliberately. Both accept a state, and setting it to draft or queue is the safe default for anything a model wrote: the post lands in the blog's drafts or its queue instead of on every follower's dashboard, Tumblr: Get Draft Posts and Tumblr: Get Queued Posts read it back, and a person approves it in the Tumblr app. Media blocks take remote URLs, so an image a workflow already produced and hosted goes straight in.

Turn engagement into a list of people

A note count is a number; Tumblr: Get Post Notes is the list of blogs behind it, with who liked, who reblogged and who replied, each with a timestamp, on a consumer-key-only connection. Tumblr: Get Blog Followers lists the blogs following one of your own and says whether you follow back, and Tumblr: Check Followed By answers the same question for one specific blog as a single boolean, which is what makes it cheap enough to run inside a loop over a prospect list. The output is a list of rows, so it flows into a Mattering database, a task, or a condition that decides whether a person ever sees it.

Read your dashboard on a schedule instead of scrolling it

Tumblr: Get Dashboard returns the posts from every blog the account follows, newest first, and takes a since_id so a scheduled run fetches only what is new rather than re-reading a window. Tumblr: Get Notifications does the same for one blog's activity feed and can be narrowed by type, which turns it into a single-purpose monitor: a new-follower alert, or a mention watcher, rather than a firehose you filter downstream. Both feed a digest, a summary in Mattering Comms, or an AI step that ranks what actually deserves a reply.

Keep a queue full and in the right order

Tumblr: Get Queued Posts reads what is waiting with each post's scheduled publish time, so a workflow can notice a queue running low and top it up with Tumblr: Create Post writing with state set to queue. Tumblr: Reorder Queued Post moves one post to promote something timely ahead of an evergreen backlog, and it is a reversible write because nothing is published until the queue reaches the post and moving it back undoes it. Tumblr: Shuffle Queue is the one to gate: nothing is published by it either, but the previous order is gone with no undo and no record of what it was.

Run a read-only Tumblr research connection with no OAuth at all

Seven nodes need nothing but the free consumer key: Tumblr: Get Blog Info, Tumblr: Get Blog Posts, Tumblr: Get Tagged Posts, Tumblr: Get Post Notes, Tumblr: Get Blog Likes, Tumblr: Get Blog Avatar and Tumblr: List Blog Pages. That is enough to mirror an archive, audit a competitor's posting cadence, or enrich a contact record with an avatar and a blog description, with a credential that structurally cannot post, like, follow or block, because the values that would let it do any of those were never entered. One caveat that is Tumblr's rather than ours: Tumblr: Get Blog Likes returns a 403 for any blog that has not turned on shared likes, which is a privacy setting and not a credential problem.

Example: An hourly tag monitor that drafts a response instead of publishing one

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

1

Trigger: an hourly schedule fires.

2

Tumblr: Get Tagged Posts reads posts from across Tumblr carrying the tag you watch. The workflow stores the newest timestamp it saw in a Mattering database and pages backwards with before until it reaches that mark, which is the correct incremental read here because this feed has no offset.

3

A condition drops anything already seen and anything below a note count worth a human's attention.

4

Tumblr: Get Post Notes reads the survivors, so the run carries the blogs already amplifying each post rather than only a count.

5

Tumblr: Get Account Limits reads what is left of the account's daily write budget, and the run stops early rather than discovering the ceiling mid-loop.

6

An AI step drafts commentary, and Tumblr: Create Text Post writes it to your blog with state set to draft, so nothing reaches a follower's dashboard yet.

7

A Mattering task opens with the source post, the notes and the draft attached, for a person to publish or discard in the Tumblr app.

Tumblr MCP questions, answered

What Tumblr operations are supported?

37 nodes on Tumblr API v2: 22 reads and 15 actions. Public reads that work on a consumer key alone are blog info, a blog's posts, the network-wide tag feed, a post's notes, a blog's public likes, avatars and custom pages. User-scoped reads are a single post in Neue Post Format, followers, following, a followed-by check, the queue, drafts, submissions, notifications, the block list, account info, account limits, the dashboard, your own likes, your account-wide following and your filtered tags. Writes are create a text post, create any post, edit, reblog, delete, like, unlike, follow, unfollow, block, unblock, mute a post's notifications, filter a tag, reorder the queue and shuffle the queue. Deliberately not built: Tumblr Communities, which is a moderation console acting on third parties rather than a workflow step; the legacy post endpoints, superseded by the Neue Post Format ones; bulk blocking, where one bad list element blocks the wrong people irreversibly; and binary media upload, because Neue Post Format media blocks take remote URLs, which is what an automation almost always has.

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.

Does Tumblr publish an official MCP server?

No, and we re-checked on 20 August 2026 rather than assuming. Tumblr's own API reference makes no mention of MCP, and nothing has been published by Tumblr or Automattic. A repository search turns up one third-party work in progress with no stars and one unrelated project matching on the name. So there is no first-party server here to compare against or to defer to, which is worth saying plainly on a page whose whole subject is a Tumblr MCP. What this connector offers instead of an assistant-only tool is the full v2 surface as workflow steps: 37 nodes that run on a schedule with nobody present, chain with every other connector in one workflow, 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.

Can I use this with just an API key, without going through OAuth?

Yes, and it is a supported configuration rather than a half-finished one. A credential holding only the OAuth consumer key runs the seven public read nodes, and Test Connection reports a read-only connection and names the three values that are missing so a half-filled form is visibly different from a deliberate choice. The other thirty nodes refuse locally with an error naming the missing fields and where to get them, rather than sending a request that would come back as an opaque 401. Add the consumer secret, the OAuth token and the token secret later and the same credential unlocks the rest without anything else changing.

Do the Tumblr tokens expire, and what do I have to refresh?

Nothing. Tumblr's OAuth 1.0a tokens never expire and never rotate, so the stored credential is a permanent set of four values and every request is signed from it locally with HMAC-SHA1. There is no token exchange, no consent round-trip at run time, and nothing written back into the secret store. That is precisely why this connector uses OAuth 1.0a rather than Tumblr's OAuth 2.0, which issues a rotating refresh token: a rotation a cold start misses would leave a dead credential with no way back, and Tumblr publishes no fallback login for that case.

Can an AI agent publish something I cannot take back?

It can, which is why nine nodes are flagged destructive and why the flag exists at all. Tumblr is public and federated: Tumblr: Create Post, Tumblr: Create Text Post and Tumblr: Reblog Post reach every follower's dashboard immediately and can be reblogged within seconds, and each reblog is an independent post on somebody else's blog holding its own copy. Tumblr: Delete Post removes your copy and cannot recall a reblog or un-send a notification. Tumblr: Edit Post replaces rather than merges, does not reach existing reblogs, and the previous version is not recoverable through the API. Tumblr: Block Blog severs any existing follow that unblocking does not restore. The practical guardrails: keep agents on the reads, write with state set to draft or queue so a person approves before anything is published, put the publishing nodes behind a condition or an approval task, and read Tumblr: Get Account Limits before a loop that writes.

What are Tumblr's rate limits?

Tumblr publishes them, so these are its numbers rather than ours. Per registered application, meaning per consumer key, 1,000 requests an hour and 5,000 a day, shared across everything using that application. Per IP address, 300 a minute, 18,000 an hour and 432,000 a day. Per account per day, 250 published posts, 250 image uploads, 200 follows, 1,000 likes, 10 new blogs, 20 videos and 60 minutes of video. The blocks endpoints carry their own ceiling of 60 requests a minute. The per-consumer-key pair is the one that actually binds, being an order of magnitude below the IP ceiling. One shape to plan around separately: every list read in this API caps limit at 20, so a workflow that wants 200 posts makes ten calls, and Tumblr: Get Blog Likes additionally refuses an offset past 1000.

Why are Unlike, Unfollow and Unblock not flagged destructive?

Because on Tumblr they genuinely restore the previous state. A like, a follow and a block are server-side state on the account, so un-liking and then liking again puts things back exactly as they were. That is different from networks built on content-addressed records, where an undo is a record delete and re-doing mints something new, and the connector deliberately does not borrow that framing here. Two honest caveats are on the nodes themselves anyway. The notification the original action sent cannot be recalled, and re-doing the action sends a new one. And Tumblr: Unblock Blog lifts the block completely but does not restore a follow the block severed, so that relationship is gone regardless. All three are still skipped by our endpoint test matrix, because Tumblr has no sandbox and each one acts on a real person's post or blog.

37 nodes waiting on the canvas

Put Tumblr 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. Tumblr is a product of its respective owner; Mattering is an independent platform.

Start Free Explore our Products