MCP Server

Connect your AI to user feedback

MCPFeedback exposes an MCP server that lets AI assistants read, create, and manage feedback items and feature requests — all via a single connection. Set up in under 2 minutes.

What can the MCP server do?

Triage feedback

List, filter, and update status on bug reports, feature requests, and UX issues across all your sites.

Manage feature requests

Create, update, and track feature requests. Analyze demand to find your most-requested features.

Post fix notes

Write resolution notes and automatically notify reporters via email when their issue is addressed.

Visual analysis

get_feedback returns screenshot and screen recording attachments as image content blocks — your AI can see what users reported.

Overview

26 tools across 5 categories. Once connected, your AI assistant can work with all of your feedback data programmatically.

Feedback

13 tools

Feature Requests

6 tools

Sites

1 tool

Crashes & Replays

3 tools

GitHub

3 tools

Quick Setup

Pick your client. The whole setup takes under 2 minutes.

Fastest path — one command, no API key

MCPFeedback's OAuth server supports Dynamic Client Registration + PKCE discovery, so most clients can connect with zero key-pasting. In Claude Code:

terminal

claude mcp add --transport http mcpfeedback https://staging.mcpfeedback.com/api/mcp

Your browser opens to sign in and choose which sites to authorize. Static API keys are still supported for headless/CI use — see the per-client tabs below.

Before you start

Create an MCPFeedback account (free trial, no credit card)
For OAuth clients (Claude Code/Connector, Cursor, ChatGPT, Gemini) you need nothing else — sign-in and site selection happen during the connect flow.
Only if your client needs a static key: go to MCP & API Keys and generate one.

1Open your terminal

2Run the one-line command below — no API key required

3Claude Code opens your browser to sign in and pick which sites to authorize

4That's it — the server is connected via OAuth

terminal

claude mcp add --transport http mcpfeedback https://staging.mcpfeedback.com/api/mcp

Zero key-pasting. MCPFeedback's OAuth server supports Dynamic Client Registration + PKCE, so Claude Code registers itself and walks you through sign-in automatically. Prefer this over a static API key. (If you must use a key instead — e.g. a headless CI box — append `--header "Authorization: Bearer YOUR_API_KEY"`.)

Share it with your team: check in a `.mcp.json`

Drop this at your repo root so every teammate's MCP-aware editor (Claude Code, Cursor) picks up the server automatically. It carries no secret — each person signs in via OAuth on first use.

.mcp.json

{
  "mcpServers": {
    "mcpfeedback": {
      "url": "https://staging.mcpfeedback.com/api/mcp"
    }
  }
}

Authentication

Two authentication methods, pick whichever your client supports:

OAuth 2.1 (Recommended)

Used by Claude Desktop Connectors, ChatGPT, and Gemini. No API key needed — sign in interactively and choose which sites to authorize.

API Key (Bearer Token)

Used by Claude Code, Cursor, and other CLI tools. Generate a key from your dashboard and pass it via the Authorization header.

API keys are account-wide and full-access: they can read and write across every site in your organization. OAuth tokens are finer-grained — during the authorization flow you choose which sites to grant and whether the token may write.

OAuth scope model — enforced server-side

mcp:read grants the read tools. A token without mcp:write is read-only — every write tool (status/severity/type updates, fix notes, deletes, bulk ops, forwarding) returns an error.
mcp:write (or mcp:admin) is required for any mutating tool.
Site-scoped tokens are enforced on every query: a token granted a subset of sites cannot see or modify feedback, crashes, feature requests, or forwards belonging to sites outside that set.
A paused/cancelled/expired account is blocked from writes on both auth methods; reads still work.

header

Authorization: Bearer mcpf_abc123...

Scoped access

Keys are scoped to your organization. Each key can access all sites and apps in your organization.

Revocable

Revoke any key instantly from your dashboard. Takes effect immediately.

Usage tracking

See when each key was last used in your API Keys dashboard.

Feedback Tools

Read, create, update, and manage user feedback items across all your sites and apps. All tools accept reference numbers (e.g. FB-123) as well as UUIDs.

list_feedback

List feedback items with free-text search plus filters for status, severity, type, and lifecycle segment. Returns { items, next_cursor } with attachment_count per item; pass next_cursor back as `cursor` for the next page.

Name	Type	Required	Description
siteId	string (UUID)	No	Filter by site ID
search	string	No	Free-text search over feedback title and details (max 200 chars)
status	string enum	No	backlog \| new \| accepted \| in_review \| in_progress \| resolved \| closed \| reopen \| rejected (overrides segment)
severity	string enum	No	low \| medium \| high
feedbackType	string enum	No	bug_report \| feature_request \| ux_issue \| general_comment
segment	string enum	No	active \| closed \| all \| deleted (default: all)
limit	number	No	Items to return, 1–100 (default: 25)
cursor	string	No	Pagination cursor (next_cursor / created_at of last item)

// Returns { items: [...], next_cursor: "2026-04-07T10:00:00Z" | null }
{
  "id": "uuid",
  "reference_number": "FB-123",
  "title": "Button unresponsive on checkout",
  "severity": "high",
  "status": "new",
  "feedback_type": "bug_report",
  "reporter_email": "user@example.com",
  "attachment_count": 2,
  "site_domain": "example.com",
  "created_at": "2026-04-07T10:00:00Z"
}

get_feedback

Get full details of a feedback item including all attachments. Accepts UUID or reference number (e.g. FB-123). Returns attachment images as inline content blocks for AI visual analysis.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number (e.g. FB-123)

// Attachments are returned as MCP image content blocks,
// so AI agents can visually analyze screenshots and
// screen recordings (as GIF) alongside the metadata.

create_feedback

Create a new feedback item with title, details, severity, and type. Generates a unique reference number automatically.

Name	Type	Required	Description
siteId	string (UUID)	Yes	The site to create feedback for
title	string	Yes	Feedback title (max 500 chars)
details	string	No	Detailed description (max 5000 chars)
severity	string enum	No	low \| medium \| high (default: medium)
reporterEmail	string (email)	Yes	Reporter email address
feedbackType	string enum	No	bug_report \| feature_request \| ux_issue \| general_comment (default: general_comment)

update_feedback_status

Update the status of a feedback item. Validates allowed status transitions. Automatically sends a resolution notification email when transitioning to 'resolved'.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number
status	string enum	Yes	backlog \| new \| accepted \| in_review \| in_progress \| resolved \| closed \| reopen \| rejected

update_feedback_type

Change the type/category of a feedback item.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number
feedbackType	string enum	Yes	bug_report \| feature_request \| ux_issue \| general_comment

update_feedback_severity

Change the severity of a feedback item.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number (e.g. FB-123)
severity	string enum	Yes	low \| medium \| high

post_fix_notes

Post a fix/resolution note on a feedback item. Does not change feedback status — call update_feedback_status separately to close or resolve. Optionally emails the reporter with the project name in the subject.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number
content	string	Yes	Fix note content (max 5000 chars)
notifyReporter	boolean	No	Send email to reporter (default: true)

list_fix_notes

List the fix/resolution notes posted on a feedback item, newest first. Read counterpart to post_fix_notes — check before posting to avoid duplicates.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number (e.g. FB-123)

get_activity

Get the activity log (status changes, assignments, replies, fix/internal notes) for one feedback item or across a whole site, newest first. Provide exactly one of feedbackId or siteId.

Name	Type	Required	Description
feedbackId	string	No	UUID or reference number (one of feedbackId / siteId)
siteId	string (UUID)	No	Read activity across all feedback for a site
limit	number	No	Rows to return, 1–200 (default: 50)

delete_feedback

Soft-delete a feedback item (recoverable) or permanently delete it.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number
permanently	boolean	No	Permanently delete (cannot be undone). Default: false

restore_feedback

Restore a soft-deleted feedback item back to the active list.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number of soft-deleted item

bulk_update_feedback_status

Update the status of up to 50 feedback items in one call. Each item is transition-validated independently; returns a per-item [{ id, ok, error }] result.

Name	Type	Required	Description
feedbackIds	string[]	Yes	1–50 UUIDs or reference numbers (e.g. FB-123)
status	string enum	Yes	Status to apply to every item

bulk_delete_feedback

Soft-delete (or permanently delete) up to 50 feedback items in one call. Returns a per-item [{ id, ok, error }] result.

Name	Type	Required	Description
feedbackIds	string[]	Yes	1–50 UUIDs or reference numbers
permanently	boolean	No	Permanently delete (cannot be undone). Default: false

Feature Request Tools

Manage the in-app feature request board. List, create, update status, and analyze demand patterns across all feature requests submitted by your users.

list_feature_requests

List feature requests across your sites. Filter by status, sort by vote count, newest, or recently updated.

Name	Type	Required	Description
site_id	string (UUID)	No	Filter by site/app ID
status	string enum	No	open \| planned \| in_progress \| shipped
sort	string enum	No	votes \| newest \| updated (default: votes)
limit	number	No	Items to return, 1–100 (default: 25)

// Example response item
{
  "id": "uuid",
  "title": "Dark mode support",
  "description": "App should respect system dark mode setting.",
  "status": "planned",
  "vote_count": 42,
  "comment_count": 8,
  "author_email": "user@example.com",
  "created_at": "2026-04-01T09:00:00Z"
}

get_feature_request

Get full details of a feature request including all comments (threaded).

Name	Type	Required	Description
id	string (UUID)	Yes	Feature request UUID

create_feature_request

Create a new feature request for a site. Appears on the in-app feature board immediately.

Name	Type	Required	Description
site_id	string (UUID)	Yes	The site to create the feature request for
title	string	Yes	Feature request title (3–200 chars)
description	string	No	Detailed description (max 5000 chars)

update_feature_request_status

Update the status of a feature request.

Name	Type	Required	Description
id	string (UUID)	Yes	Feature request UUID
status	string enum	Yes	open \| planned \| in_progress \| shipped

comment_on_feature_request

Post an official owner comment on a feature request. Marked as an owner reply; can reply to an existing thread. Blocked when the site disables comments.

Name	Type	Required	Description
id	string (UUID)	Yes	Feature request UUID
content	string	Yes	Comment body (1–2000 chars)
parentCommentId	string (UUID)	No	Reply to an existing comment thread

analyze_demand

Analyze feature request demand across your sites. Returns top voted requests, requests grouped by status, and trending requests from the last 7 days.

Name	Type	Required	Description
site_id	string (UUID)	No	Scope analysis to a specific site
limit	number	No	Number of top requests to return (default: 10, max: 50)

// Example response
{
  "top_voted": [
    { "title": "Dark mode", "vote_count": 42, "status": "planned" },
    { "title": "CSV export", "vote_count": 31, "status": "open" }
  ],
  "by_status": { "open": [...], "planned": [...] },
  "trending": [...]
}

Name	Type	Required	Description
includeInactive	boolean	No	Include inactive sites (default: false)

Crash & Replay Tools

Inspect mobile crash reports (crashes, ANRs, hangs) and session replays captured by the native SDKs — grouped by fingerprint, with stack traces, device info, network logs, and per-frame replay images.

list_crashes

List mobile crash reports (crashes, ANRs, hangs) for a site or across the account, grouped by fingerprint and sorted by most recently seen.

Name	Type	Required	Description
siteId	string (UUID)	No	Scope to a specific site (default: all sites)
type	string enum	No	crash \| anr \| hang
platform	string enum	No	flutter \| ios \| android
search	string	No	Free-text search over the exception class
limit	number	No	Crash groups to return, 1–100 (default: 25)

get_crash

Get full details of a single crash report: stack trace, threads, device info, network logs, plus any linked session-replay IDs.

Name	Type	Required	Description
crashId	string (UUID)	Yes	The crash report ID

get_replay

Get a session replay: user actions, network events, and per-frame screenshot URLs (short-lived signed URLs).

Name	Type	Required	Description
sessionId	string (UUID)	Yes	The replay session ID

GitHub Tools

When a site has the GitHub App connected, forward feedback to issues and track the forwarding outbox. Writes (forward_to_github) require a connected integration with a configured repository.

get_github_issue_url

Get the GitHub issue linked to a feedback item. Returns { url, number, state } or null when not linked.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number (e.g. FB-123)

forward_to_github

Manually enqueue a feedback item for forwarding to GitHub. Requires the site's GitHub integration to be connected with a configured repository.

Name	Type	Required	Description
feedbackId	string	Yes	UUID or reference number

list_pending_github_forwards

List GitHub forward outbox rows in pending/failed/dead state for this account, so an agent can see what still needs to reach GitHub.

Name	Type	Required	Description
limit	number	No	Rows to return, 1–100 (default: 25)

Try these prompts

Once connected, try asking your AI assistant:

“Show me all unresolved bug reports”

“What are the top 10 most-voted feature requests?”

“Mark FB-12 as resolved”

“Create a feature request: CSV export for feedback data”

“Analyze feature request demand and find trending requests”

“Post a fix note on FB-12 saying we fixed the layout bug”

“List all high-severity feedback from the last week”

“What feedback types are most common on my site?”

Common questions

Things that trip people up the first time:

Server details

Server URL	`https://staging.mcpfeedback.com/api/mcp`
Transport	Streamable HTTP (JSON-RPC)
Auth	OAuth 2.1 (PKCE + Dynamic Client Registration) or Bearer API key
OAuth Discovery	`https://staging.mcpfeedback.com/.well-known/oauth-protected-resource`
Tools	26 tools across 5 categories: feedback, feature requests, sites, crashes & replays, GitHub

Ready to connect?

Create your organization and generate an API key to get started.

Get started free Go to API Keys