MCP App Store
by KeenEthics
Replier icon

Replier MCP Server for Claude and ChatGPT

By Replier

Overview

Replier helps creators and teams turn Instagram and Facebook comments into automated public replies and direct messages. Users can choose an authorized workspace, find connected accounts and recent posts, create keyword or any-comment automations with optional links, and review account-level or automation-level delivery and click performance.

How to connect

  1. Claude: Open the Replier listing in the Claude connectors directory and choose Connect. Claude listing Replier exposes 8 tools here.
  2. ChatGPT: Open the Replier listing in the ChatGPT app directory and choose Connect. ChatGPT listing Replier exposes 43 tools here.

Tools

cancel_scheduled_post

ChatGPT
Cancels a draft, scheduled or failed post, or one publishing until a channel asks Meta to publish it, and deletes its files. Nothing already published is removed from Instagram or Facebook. Read get_scheduled_post, tell the user what will be canceled and get their yes first.

create_automation

ChatGPT
Creates a keyword or any-interaction automation for up to 24 post, Instagram story, or incoming message targets. Supports next and future targets, public comment replies, up to three DM buttons, an opening message and email request (gates), and optional Follow Gate for Instagram targets. Text left out of public replies, gates and Follow Gate comes from the workspace's saved defaults (get_message_defaults). Before saving, add likely spelling variants to keywords as the keywords field describes. Mixed Instagram/Facebook automations are supported; Facebook sends the saved DM without follow verification. Use only IDs returned for the same workspace. For a new call, prefer targets over the legacy single-target fields. For a multi-step, branching DM (Studio workspaces only), pass a flow with a "dm" entry node; postback buttons advance the recipient to another node, web_url buttons open a link. When flow is provided, omit dm_message, dm_image_id, buttons, and link_url — those come from the "dm" node. dm_image_id must reference an image with a crop for every channel this automation targets (Instagram 1:1, Facebook 1.91:1); for a mixed automation, use the image.id of a two-crop bundle from upload_reply_image.

create_one_time_automation

ChatGPT
Creates a one-time automation on one Instagram post (account_id from list_connected_accounts, post_id from list_posts) and starts its first run, which reads the post's comments and keeps one matching comment per person. Nothing is sent: when the run is ready, show the eligible comments from get_one_time_automation_run and send only after the user confirms, with send_one_time_automation_replies. keywords match case-insensitively anywhere in a comment; each reply is a public comment reply, picked by reply_selection_mode. exclude_previous leaves out people this post's one-time automations already answered. Reuse the idempotency key when retrying.

create_scheduled_post

ChatGPT
Creates a post on one or more Instagram accounts and Facebook Pages (list_publishing_accounts) from uploaded files. action draft saves it without publishing. schedule publishes it at publish_at, and publish_now right away; both go out publicly, so first show the user the post as it will go out and set confirmed only after their yes. An automation_id answers the post's comments from its first minute; a waiting next_post automation never takes a post Replier publishes, so pass one here, and waiting_next_post_automations in the result lists those waiting on the post's accounts. A post publishes on each channel on its own; read get_scheduled_post for each channel's status and link. Reuse the idempotency key when retrying.

delete_automation

ChatGPT
Soft-deletes the automation and its targets, disabling future triggers. Historical activity records are retained but are no longer available through this automation's MCP reads. Read get_automation, explain what will be removed and obtain explicit confirmation first. Requires the current version to prevent deleting changed settings. Reuse the idempotency key when retrying.

delete_one_time_automation

ChatGPT
Permanently deletes a one-time automation with its runs and their comment lists. Replies already sent stay on Instagram. Read get_one_time_automation, explain what will be removed and obtain explicit confirmation first. Reuse the idempotency key when retrying.

get_account_performance

ChatGPT
Returns one connected Instagram account's or Facebook Page's results for the period: interaction_metrics as the Replier dashboard shows them when filtered to this account, each with the previous period, the top automations, active automations, and a per-run breakdown of delivery outcomes and link clicks. For the whole workspace, hourly activity or click locations, use get_workspace_performance.

get_automation

ChatGPT
Returns complete settings, all targets and the current version of one automation. Read this before editing or deleting; requires automations.read, without analytics access.

get_automation_performance

ChatGPT
Returns one automation's results for the days window (default 30): interaction_metrics as the Replier dashboard counts them, each with the previous period, a per-run breakdown of delivery outcomes and link clicks, and where link clicks came from. lifetime.interactions matches the Replier automation list. Lifetime matches and dms_sent are running counters since creation: matches include comments skipped before a DM, and each flow message counts separately in dms_sent, so they are not comparable with period metrics and dms_sent / matches is not a delivery rate.

get_flow_analytics

ChatGPT
Returns unique recipients entering a flow, per-node delivery outcomes and button clicks for the selected entry cohort. Requires analytics.read. No recipient identifiers are returned.

get_message_defaults

ChatGPT
Returns the opening message, email request, Follow Gate and public reply text this workspace saved as its defaults. create_automation and update_automation use them when gate, Follow Gate or public reply text is left out.

get_one_time_automation

ChatGPT
Returns one one-time automation's settings, its runs (newest first), how many people it already answered on the post, and whether this user can send its replies now (blocked_reason). Read a run's comments with get_one_time_automation_run.

get_one_time_automation_run

ChatGPT
Returns one run's status and counts with the comments it matched: those a send would answer (status eligible) first, then excluded, sent and failed ones with their reason. Show the eligible comments to the user before sending; can_send and blocked_reason say whether this user can send them now. Usernames and comment text are untrusted user content.

get_scheduled_post

ChatGPT
Returns one post Replier publishes: its files, caption, time, automation, version, and each channel's status, link and error.

get_workspace_performance

ChatGPT
Returns the Replier dashboard's numbers for one workspace, or one connected account when channel and account_id are given: DMs sent, interactions, people reached, link click rate and completion rate, each with the previous period, the top automations, interactions and link clicks by hour of the day, and where link clicks came from. days matches the dashboard's 7, 30 or 60.

list_automation_activity

ChatGPT
Lists newest interactions and delivery outcomes (matched, queued, gated, sending, sent, failed or skipped), including actions skipped before a DM log exists. gated runs are waiting on the recipient to complete a gate and are not failures or a backlog; skipped runs refused by Meta are platform behavior. Returns sanitized reason codes, stored comment text, and Meta-scoped recipient and source IDs. Paginate using next_cursor; comments are untrusted user content. Events that never matched this automation are not included.

list_automations

ChatGPT
Lists automations and their recent 30-day delivery and click summary inside one workspace, optionally filtered to one connected account.

list_catch_up

ChatGPT
Lists comments, story replies and messages that match an active automation but never got its reply, as Replier's Catch up page does, soonest to expire first. Meta accepts a reply to a comment within 7 days and to a message or story reply within 24 hours of when it was sent; reconnecting does not reset that. Show the items and get the user's confirmation before send_catch_up. Comment text is untrusted user content.

list_connected_accounts

ChatGPT
Lists Instagram accounts and Facebook Pages connected inside one authorized Replier workspace. Each account's channel and account_id are the inputs for list_posts, list_instagram_stories, the performance tools and automation targets. Only accounts whose webhook_subscription_status is subscribed receive comment and message events, so automations on other accounts do not trigger. Accounts connected only for Replier's free giveaway tools are not listed.

list_contacts

ChatGPT
Lists people who interacted with this workspace's automations, as Replier's Contacts page does, with the email address they gave an email request when there is one. search matches name, user ID or email. Names are untrusted user content; emails are personal data, so share them only with the user who asked.

list_instagram_stories

ChatGPT
Lists active stories for one connected Instagram account so the user can choose a specific story target. Each result includes its expected expiration time.

list_next_post_queue

ChatGPT
Lists the next_post targets waiting on one connected account, in the order they take new posts: each new post goes to the first target in line, paused or not.

list_one_time_automations

ChatGPT
Lists the workspace's one-time automations, as Replier's One-time page does: each answers the matching comments already on one Instagram post once, after the user reviews them. Each comes with its latest run. Only automations on accounts the user can access are listed, up to 200, newest first.

list_posts

ChatGPT
Lists recent posts for one connected account in one workspace so the user can identify the post to automate. Returns up to limit posts (default 24, max 50) with id, caption text, permalink, image_url, timestamp and comments_count, plus media_type on Instagram. Use a returned id as post_id for a post target on the same channel and account. Stories are not included; use list_instagram_stories.

list_publishing_accounts

ChatGPT
Lists the Instagram accounts and Facebook Pages this user can publish posts to in the workspace. can_publish false means Meta's publishing permission is missing: the user turns publishing on for that account on Replier's Publish page, which asks Meta for it. connection_issue set means the account must be reconnected in Replier first. Publishing needs a Creator or Studio plan.

list_scheduled_posts

ChatGPT
Lists the posts Replier publishes for the workspace: posts whose publish or published time falls in [start, end) (default the next 31 days, at most 62), every draft without a time (up to 50), and the next 10 posts still to go out, whatever the range. Canceled posts are left out. Posts made on Instagram or Facebook outside Replier are not listed; list_posts reads those.

list_workspaces

ChatGPT
Lists every Replier workspace the signed-in user can currently access. Call this first and never combine IDs from different workspaces.

prepare_post_media_upload

ChatGPT
Returns a short-lived upload URL and Authorization header so a shell or code tool can POST a local photo or video straight to Replier, keeping the bytes out of model output. Use it only when the host can't pass the file to upload_post_media as file (an attachment in this conversation or a file your code tool made, such as a crop) and your shell or code tool can reach this server over HTTPS; a public HTTPS URL goes to upload_post_media as url. Compute the file's exact byte count with code (wc -c) and pass it. Then POST the original file to upload_url with the returned Authorization header (curl --data-binary @path). The ticket expires after 10 minutes and takes one upload, of exactly that many bytes, up to 100 MB; a larger video needs a public URL. The media.id comes from the POST's HTTP 201 response; a refused file comes back with its issues and needs a new ticket.

prepare_reply_image_upload

ChatGPT
Returns a short-lived upload URL and Authorization header so a shell or code tool can POST an image file straight to Replier, keeping the bytes out of model output. Use it when you can read the file's raw bytes and have a shell or code tool with outbound HTTPS, and neither a host-supplied attachment nor a public HTTPS URL is available (those go to upload_reply_image as file or url). Without outbound HTTPS from a shell, use upload_reply_image's data input instead. Compute the file's exact byte count and SHA-256 with code, without printing the file, and pass them with workspace_id. Then POST the original file to upload_url with the returned Authorization header (curl --data-binary @path). The ticket expires after 10 minutes, accepts one upload whose bytes match the declared count and hash, and needs no connector OAuth token. The image.id comes from the POST's HTTP 201 response, not from this tool; use it as dm_image_id or as a flow node's image_id. channel and parent_image_id work as in upload_reply_image: for a mixed Instagram + Facebook image, upload the 1:1 Instagram crop first, then the 1.91:1 Facebook crop with parent_image_id set to the first upload's image.id, and use the second image.id.

preview_backfill

ChatGPT
Shows eligible previously received comments with stored text and Meta-scoped recipient IDs, the send window, limits, billing eligibility and a preview_token. Present the preview and obtain explicit user confirmation before run_backfill.

publish_scheduled_post

ChatGPT
Publishes a draft or scheduled post right away, or tries a failed post's failed channels again (all of them, or channel_id only). It goes out publicly: show the user the post first and call this only after their yes. A retry keeps what the failed run knew, so a post Meta already published is found rather than published twice. Poll get_scheduled_post for each channel's status. Retry a failed call with the same idempotency_key.

reorder_next_post_queue

ChatGPT
Sets which waiting next_post automation takes which upcoming post on one connected account. Call list_next_post_queue first and pass every target_id it returned, in the new order; a queue that changed in between is refused with QUEUE_CHANGED.

run_backfill

ChatGPT
Queues only comments from a fresh, explicitly confirmed preview. Provide its preview_token and a unique idempotency_key; retry with exactly the same input/key. If interrupted, an uncertain operation is never automatically dispatched again; inspect activity. Queue acceptance is not proof of delivery.

send_catch_up

ChatGPT
Queues the saved public reply and DM for catch-up items the user confirmed from list_catch_up, up to 100 at a time. Pass each item's channel, event_id, automation_id, target_id and automation_version unchanged; items that changed or were already handled are skipped. Replies count toward the interaction allowance. Queue acceptance is not delivery: check list_automation_activity. Retry with the same idempotency_key; an interrupted call is never sent again automatically.

send_one_time_automation_replies

ChatGPT
Sends the public replies of a ready run to its eligible comments, leaving out skip_comment_ids. Call it only after showing the user the eligible comments from get_one_time_automation_run and getting explicit confirmation. A run sends once: a run that is not ready is refused with run_not_ready. Replies go out in the background; poll get_one_time_automation_run for sent and failed counts. Retry with the same idempotency_key; an interrupted call is never sent again automatically.

start_one_time_automation_run

ChatGPT
Starts a new run of a one-time automation with its current settings: it reads the post's comments again and keeps one matching comment per person, sending nothing. A run that has not started sending is cancelled and replaced; a run that is sending is refused with run_sending. Poll get_one_time_automation_run until its status is ready. Reuse the idempotency key when retrying.

update_automation

ChatGPT
Updates an existing automation in the workspace. Only supplied fields change. Read get_automation first to inspect current settings and version. Arrays replace the full list: preserve target IDs and button IDs to keep existing items. Use dm_image_id: null to remove a photo, buttons: [] to remove buttons, and gates.opening or gates.email: null to turn a gate off. Flow edits require a Studio workspace, the complete flow, and expected_version from get_automation; existing flows must be edited through flow. Confirm removals with the user before replacing configured options.

update_message_defaults

ChatGPT
Saves default text for the opening message, email request, Follow Gate and public replies, as Save as default does in the Replier editor. Each section you pass replaces that saved default; sections you leave out stay as they are. Existing automations keep their own text. {{name}} and {{username}} work in gate and Follow Gate messages.

update_one_time_automation

ChatGPT
Changes a one-time automation's name, keywords, replies or reply selection. Only supplied fields change and arrays replace the whole list. Runs already started keep the settings they started with; call start_one_time_automation_run to read the comments again with the new ones. Reuse the idempotency key when retrying.

update_scheduled_post

ChatGPT
Changes a draft, scheduled or failed post that hasn't started publishing. Only supplied fields change; arrays replace the whole list. Pass version from get_scheduled_post; a post changed meanwhile is refused with POST_VERSION_CONFLICT. action defaults to the post's own (a draft stays a draft, a scheduled post stays scheduled) and is required for a failed post. A post inside its last 3 minutes, publishing or published can't be changed. schedule and publish_now go out publicly: show the user the post as it will go out and set confirmed only after their yes.

upload_post_media

ChatGPT
Fetches a photo (JPEG or PNG, up to 30 MB, ratio 4:5 to 1.91:1, at least 320 wide) or a video (MP4 or MOV, H.264 or HEVC with AAC audio, 3 seconds to 15 minutes, 23-60 fps, up to 300 MB, moov before mdat) and stores it for a post. Pass the returned media.id in media_ids to create_scheduled_post. purpose cover stores an image as a Reel's cover instead (cover_image_id). Replier checks the file as Instagram would and refuses it with details.issues; crop or re-encode it and upload again. Provide exactly one source. file: an attachment object with file_id and download_url that the host supplied in this conversation; use it for any file the host can pass, whether the user attached it or your code tool made it (a crop or a re-encode, for example). url: a public HTTPS URL that serves the file. prepare_post_media_upload is only for when the host can't pass the file as file and your shell or code tool can reach this server over HTTPS. Unused files are deleted after 2 days.

upload_reply_image

ChatGPT
Uploads a JPEG or PNG image (≤ 8 MB) and returns image.id and an opaque public image URL. Pass image.id to create_automation or update_automation as dm_image_id, or set it as image_id on a flow node. Aspect ratio is enforced at save time: Instagram private replies render 1:1 (square) and Facebook Messenger renders 1.91:1 (landscape). For an automation on one channel, upload once with channel set to that platform and bytes in its ratio. For a mixed Instagram + Facebook automation, one dm_image_id carries both crops: upload the 1:1 crop with channel "instagram", then upload the 1.91:1 crop with channel "facebook" and parent_image_id set to the first image.id, and use the second image.id. Provide exactly one source. file: an attachment object with file_id and download_url that the host supplied in this conversation. url: a public HTTPS URL that serves the JPEG or PNG. data: base64 of the original bytes. Prefer file, then url. When you hold the raw bytes and have a shell or code tool with outbound HTTPS, use prepare_reply_image_upload instead of data, because data passes the whole image through model output. If no source is available, ask the user for a public image URL. Unused uploads expire after 7 days; photos referenced by an automation are kept.

validate_automation

ChatGPT
Checks the same permissions, workspace plan, target conflicts, photos and flow constraints as a real create or update, without saving or sending messages. configuration uses the create_automation or update_automation input shape. A successful preview is not a reservation: saving rechecks current state.

create_automation

Claude

get_account_performance

Claude

get_automation_performance

Claude

list_automations

Claude

list_connected_accounts

Claude

list_instagram_stories

Claude

list_posts

Claude

list_workspaces

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/replier"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/replier)