MCP tools

These are the tool names your agent sees when it connects, so you can match them to what it shows in a session. You never call them yourself: you ask in plain words and the agent picks the tools. How to connect: MCP server.

​

The order an agent follows

​

Channels

​

integrationList

Lists the channels the agent can post to. Channels that are turned off, need to be reconnected or were not finished connecting are left out. Returns output, a list of channels: id, name (the name you gave the channel in PostQueen, or the account’s name on the network), platform (the network, such as bluesky or x), picture, display (the account handle), type and customer (id and name, when the channel belongs to one).

​

groupList

Lists your customers, the groups your channels are sorted into. Takes no input. Returns output, a list of id and name.

​

integrationSchema

Returns a network’s rules and the settings a post to it needs. The agent calls it before it writes a post for a network. Returns output with:

​

triggerTool

Runs one of a network’s helpers to look up a value a setting needs, such as a list of WordPress post types or DEV tags. Returns output, whatever the network answered: usually a list of id and name pairs. If the channel’s sign-in has expired, the tool tries to renew it. When that fails, the channel is marked to be reconnected and the tool says so. A channel that was deleted answers Integration not found, use integrationList to get a valid integration id.

​

Posts

​

integrationSchedulePostTool

Creates posts: a draft, a post scheduled for a time, or a post that goes out now. PostQueen checks each post the way the app does before it saves anything. Returns output, a list of postId and integration for each post it created. When a post fails the checks, output is errors with the reason, and nothing is created. The agent fixes the post and calls the tool again. An integrationId that is not a channel of your workspace, or a channel that was deleted, returns errors: Integration not found.

• Settings and length are checked for scheduled posts and posts that go out now. Over MCP, a draft is saved even when a required setting is missing, so add it before the draft is scheduled. The CLI and the API check a draft’s settings too.
• Attachments must be media paths that end in .png, .jpg, .jpeg, .gif, .webp or .mp4. Bring a file in with an upload tool first.
• On WordPress, Listmonk and an X Article, pictures go inside the content as <img src="..." alt="..."> where they should appear. The src is the path of a picture in your media library. integrationSchema says so in the network’s rules.
• What extra items in postsAndComments become depends on the network: see comments and threads.

​

postsListTool

Lists your posts between two dates, whatever their state. Returns output.posts, each with id, publishDate (UTC), state (QUEUE, DRAFT, PUBLISHED or ERROR), error (why a post with the ERROR state failed, null otherwise), content, settings, group, integrationId, platform and integrationName.

​

postReadTool

Reads one post in full. Returns output with the post’s id, group, integrationId, platform, integrationName, state, publishDate and settings, and posts: the post and each comment or thread item, with its id, content, delay and attachments. An unknown ID returns error.

​

postSettingsTool

Changes the network settings of a post that has not gone out. The text and the publish time stay as they are. Returns output with postId and publishDate, or errors. It works on drafts and on scheduled posts whose time has not passed. The tool tells MCP clients that it changes data that is already saved, so a client that asks before such changes asks you to approve it.

​

Analytics

The three analytics tools read your published posts on the networks that report numbers: X, Facebook Page, Instagram, LinkedIn Page, YouTube, Threads, Pinterest, TikTok and Google Business Profile. Analytics lists what each one reports.

• The window (7, 30 or 90 days) picks which posts count, by publish date. The numbers are each post’s totals so far, not only what happened inside the window.
• null means the network does not report that number. Facebook gives no comment counts, Pinterest gives impressions only, and Google Business Profile gives no numbers per post.
• syncing is true while PostQueen is still fetching the latest numbers.

​

analyticsSummaryTool

Totals for the posts you published in the window. Returns output with syncing, date, posts (how many), reactions, comments, impressions and notes.

​

analyticsPostsTool

Ranks your posts, or searches them, by how they did. Returns output with total, page, limit, posts, and topReactions and topComments (the five strongest in the window, whatever the sort). Each post has id, content, publishDate, releaseURL, integrationId, platform, channelName, impressions, reactions, comments, shares and engagementRate.

​

analyticsPostTool

The numbers for one published post. Returns output.post with the same fields as a post in analyticsPostsTool, or output.error: not_found, not_published, unsupported, or missing_release when PostQueen does not know the post’s ID on the network yet. Connect the post to its published copy in the app, or with postqueen posts:connect.

​

Uploads

A post takes media from your PostQueen media library. These tools put a file there and return its path for attachments.

​

uploadFromUrlTool

Brings an image or video in from a public http or https link. Returns the file’s id, path and status, or error. An image is ready at once. An MP4 comes back as processing while PostQueen prepares it, and its path can change when it is done. A MOV is converted to MP4 first, and the tool waits for that, up to about four minutes. The tool cannot reach a file on your computer or on a private address. Upload those with the CLI (postqueen upload) or the API (Upload a file), then find them with mediaListTool.

​

mediaListTool

Lists what is already in your media library, newest first, 18 to a page. The agent uses it to reuse a file instead of uploading it again. Videos that are still being converted are left out. Returns pages, the number of pages, and output, each file with id, name, originalName, path and createdAt. The id and path go into a post as they are.

​

uploadWidgetTool

Opens an Upload Media panel in the chat, where you add photos or videos from your device. It works in apps that can show MCP panels, such as Claude, Claude Cowork and ChatGPT. Elsewhere no panel appears: upload a file from your computer with the CLI or the API, or use uploadFromUrlTool for a public link. Takes no input. Returns a sessionId. What you upload through the panel is reported in the chat, and uploadWidgetStatusTool lists it.

​

uploadWidgetTicketTool

Used by the upload panel itself, not by your agent. It gets a short-lived upload ticket for the panel, so the credential never appears in the chat. Returns a ticket, or error.

​

uploadWidgetStatusTool

Lists what you uploaded through the panel in that session, not the whole media library. Returns media, each with id, path and status. An empty list means nothing was uploaded yet. A file that is processing is still being prepared: the agent waits about 10 seconds and asks again. A ready file can go into a post.

​

Images and video

Creating an image or a video is paid from your workspace’s credits. What it costs depends on the settings you choose. When the balance is too low, the tool answers Not enough credits: this costs N and the balance is N. and nothing is made.

​

generateImageTool

Creates an image and puts it in your media library. Returns the image’s id, path, alt and thumbnail, or error. An image that could not be made reads Image generation failed:, the reason, and No credits were used.

​

generateVideoOptions

Lists the video generators your agent can use. The agent calls it before it starts a video. An empty list means none is available. Returns video, one entry per generator: The two generators take these inputs:

​

videoFunctionTool

Looks up a value a video generator needs, with one of the helpers generateVideoOptions listed under tools. The only one today is loadVoices on image-text-slides. Returns the helper’s answer. loadVoices returns voices, each with an id (the voice input), a name and a preview_url with a sample. A wrong call names what you can use instead:

​

generateVideoTool

Starts a video and puts it in your media library when it is done. It takes minutes, so the tool returns a job ID at once. The price depends on the generator and its inputs, and a video that fails gives its credits back. Returns a jobId, or error when the video could not start. No credits are used then. An identifier that is not a generator is refused before anything else, with There is no video generator "<identifier>". Use one of these identifiers: ... No credits were used. Use one of the identifiers it lists. A balance that is too low answers Not enough credits: ... as above. Every other error reads Video generation failed:, the reason, and No credits were used. The reasons you are most likely to see:

​

videoStatusTool

Checks a video job. Returns status:

• pending: the video is still being made. The agent waits about 30 seconds and asks again.
• completed, with the media’s id and the video’s url. The url goes into attachments as it is.
• failed, with the error, such as AI generation failed, please try again later. The credits are given back.

A job ID of another workspace, or one PostQueen no longer keeps, returns error: Video job lookup failed: Video job not found. A video that finished is in your media library even when its job is gone.

​

Clipping

Clipping turns a long YouTube video into short vertical clips with captions, saves them in your media library and can add them to your channels as drafts. It uses your plan’s clipping minutes: one for each started minute of the video. How it works: Clipping.

​

clippingTool

Starts a clipping. Before it calls the tool, the agent asks you how the horizontal video should fill the vertical clip, and waits for your answer. Returns a clippingId, or error. A clipping takes several minutes, so the tool only starts it. In apps that can show MCP panels, a Video Clipping panel follows its steps, shows each clip with Open and Copy link, and tells the chat when the clips are ready.

​

clippingStatusTool

Checks a clipping. While the clipping runs, the tool waits up to 25 seconds for something to change before it answers, so the agent can ask again straight away. Returns status: pending while it runs, completed, or failed. With it come step (analysing, transcribing, picking or rendering while it runs), title (the video’s title), error, and clips, each with id, title, content (the post text), status, url (the clip’s video, ready to attach to a post), thumbnail and error. A completed clipping can still carry failed clips. A failed one made no clip, and its minutes were given back.

​

clippingWidgetTicketTool

Used by the Video Clipping panel itself, not by your agent. It gets a short-lived ticket to read one clipping’s progress, so the credential never appears in the chat. Returns a ticket, or error.

​

Agent

​

ask_postqueen

Sends a request in plain words to PostQueen’s own agent, which picks and runs the other tools itself. Before it schedules a new post, it writes out the draft and waits for a yes. It is on the key address and the Bearer header only. Returns the agent’s answer. The agent’s work is paid from your workspace’s credits by the tokens it uses, and any image or video it makes at its own price. With a balance of zero or less, the tool is refused with Not enough credits: ....

​

What the tools cannot do

​

Next steps