Feedbucket MCP

Table of Contents

• What is Feedbucket MCP?
• Tool List
• • projects-list
  • projects-create
  • project-archive
  • project-get
  • project-update
  • feedback-list
  • feedback-get
  • feedback-update
  • attachment-create
  • comment-create
  • comment-update
  • comment-delete
  • feedback-get-console-log
• How to Connect
• • Cursor
  • VS Code
  • Claude Code
  • Claude Desktop
  • Claude.ai
  • ChatGPT
  • Codex
• Access Tokens
• • Create a token
  • Use a token
  • Connected apps

What is Feedbucket MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants — Cursor, Claude, ChatGPT, Codex, and others — securely call tools exposed by external services. Instead of copying data back and forth by hand, the assistant can query and act on your product data in place.
Feedbucket MCP connects those assistants to your Feedbucket account. From the editor or chat, an assistant can:

• List, create, archive, and update projects
• Read feedback (including annotated screenshots and browser console logs)
• Resolve or reopen feedback
• Comment on feedback threads, including file attachments
• Edit or delete comments
• Change project settings (widget visibility, tags, language, and more)

The server is a Streamable HTTP MCP endpoint at:

https://dashboard.feedbucket.app/mcp

All tools operate on the authenticated user's current team. Authentication uses either a personal access token (Bearer header) or OAuth (browser sign-in), depending on the client.

Tool List

projects-list

Tool name: projects-list

What it does: List projects for the authenticated user's current team. Read-only and idempotent. Results are ordered by creation date (newest first).

Input schema

Output

{
  "projects": [
    {
      "id": 1,
      "name": "Acme Marketing Site",
      "url": "https://example.com",
      "imageUrl": "https://...",
      "type": "website",
      "key": "abc123",
      "lastActivityAt": "2026-07-01T12:00:00+00:00",
      "lastFeedbackAt": "2026-07-01T11:30:00+00:00",
      "createdAt": "2026-01-15T09:00:00+00:00",
      "archived": false
    }
  ]
}

Errors

• No team found for the authenticated user.

projects-create

Tool name: projects-create

What it does: Create a new project for the authenticated user's current team.

Input schema

Output

{
  "project": {
    "id": 1,
    "name": "Acme Marketing Site",
    "url": "https://example.com",
    "imageUrl": null,
    "type": "website",
    "key": "abc123",
    "lastActivityAt": null,
    "lastFeedbackAt": null,
    "createdAt": "2026-07-17T10:00:00+00:00",
    "archived": false
  }
}

Errors

• No team found for the authenticated user.
• Project limit reached for your plan.

project-archive

Tool name: project-archive

What it does: Archive or restore a project by ID for the authenticated user's current team. By default the project is archived; set archived to false to restore it (subject to your plan's active project limit).

Input schema

Output

{
  "success": true
}

Errors

• No team found for the authenticated user.
• Project not found.
• You do not have access to this project.
• Active project limit reached for your teams plan. (on restore)

project-get

Tool name: project-get

What it does: Get full details for a single project by ID, including all settings. Read-only and idempotent.

Input schema

Output

{
  "project": {
    "id": 1,
    "name": "Acme Marketing Site",
    "url": "https://example.com",
    "imageUrl": "https://...",
    "type": "website",
    "key": "abc123",
    "lastActivityAt": "2026-07-01T12:00:00+00:00",
    "lastFeedbackAt": "2026-07-01T11:30:00+00:00",
    "createdAt": "2026-01-15T09:00:00+00:00",
    "archived": false,
    "settings": {
      "visibility": "all",
      "theme": "default",
      "widgetVisible": true,
      "feedbackSubmissionOpen": true,
      "controlsPlacement": "bottom-right",
      "showPins": "always",
      "enabledFeatures": ["record", "view_feedback", "comment", "resolve", "tag", "mention"],
      "pinColor": "theme_primary",
      "onboardingEnabled": true,
      "onboardingWelcomeHeading": null,
      "onboardingWelcomeBody": null,
      "language": "en",
      "embedImages": false,
      "embedCSS": false,
      "queryTriggerSecret": null,
      "basicAuth": false,
      "basicAuthUsername": null,
      "basicAuthPassword": null,
      "notifications": {
        "enabledFor": ["comment", "resolve"]
      },
      "textOverrides": {
        "feedbackDescriptionPlaceholder": null
      },
      "tags": [
        {
          "name": "Bug",
          "syncsWith": null
        },
        {
          "name": "Design",
          "syncsWith": "jira"
        }
      ]
    }
  }
}

Errors

• No team found for the authenticated user.
• Project not found.
• You do not have access to this project.

project-update

Tool name: project-update

What it does: Update a project's name, URL, and/or settings. Only provided fields are changed. At least one of name, url, or settings must be provided.

Input schema

settings properties (all optional):

Some settings (for example enabledFeatures, custom onboarding welcome text, and textOverrides) are plan-gated and are silently ignored if your plan does not include them.

Output

Same full project object as project-get, including the updated settings.

Errors

• No team found for the authenticated user.
• Project not found.
• You do not have access to this project.
• At least one of name, url, or settings must be provided.
• Validation errors for invalid settings values

feedback-list

Tool name: feedback-list

What it does: List feedback for a project. Read-only and idempotent. Results are ordered by creation date (newest first). Defaults to open feedback only.

Input schema

Date filters

• Values may be a date (2026-08-01) or a full datetime (2026-08-01T14:30:00+02:00).
• createdAfter / createdBefore filter only on when the feedback was submitted.
• changedAfter / changedBefore match feedback if any of these happened in the range:
• • the feedback was created
  • a comment was added
  • an activity happened (title/description/tags/members-only change, resolve, reopen)
• Filters combine with AND. Example: status: open + changedAfter: <timestamp> returns open feedback that was created or changed since that time.
• Date-only values are treated as midnight at the start of that day.

Example: to sync only what moved since the last automation run, call feedback-list with status: open and changedAfter set to the timestamp of the previous run.

Output

{
  "feedback": [
    {
      "id": 42,
      "title": "Logo is blurry on mobile",
      "description": "The logo looks pixelated on iPhone.",
      "descriptionTruncated": false,
      "tags": ["Bug", "Design"],
      "status": "open",
      "reporter": {
        "type": "reporter",
        "name": "Jane Client",
        "email": "jane@client.com"
      },
      "membersOnly": false,
      "page": "https://example.com/about",
      "createdAt": "2026-07-10T14:22:00+00:00",
      "changedAt": "2026-07-10T15:00:00+00:00"
    }
  ]
}

Output details

changedAt is always returned (not only when date filters are used). It is the latest of:

• createdAt
• the latest comment time
• the latest activity time (title/tags/resolve/etc.)

If nothing has happened since creation, changedAt equals createdAt.

When reporter.type is member, an id field is also included on the reporter object.

description is a preview of the feedback text, truncated to 500 characters (with ... when cut). descriptionTruncated is true when the preview was truncated — call feedback-get for the full description.

Errors

• No team found for the authenticated user.
• Project not found.
• You do not have access to this project.

feedback-get

Tool name: feedback-get

What it does: Get full details for a single feedback item. Read-only and idempotent. When the feedback type is screenshot and the annotated image is ready and ≤ 3 MB, the screenshot is also returned inline as an image in the response. If the image is still generating, too large, or missing, only the JSON payload is returned (resource may be "generating" while processing).

Input schema

Output

{
  "feedback": {
    "id": 42,
    "title": "Logo is blurry on mobile",
    "description": "The logo looks pixelated on iPhone.",
    "tags": ["Bug", "Design"],
    "attachments": [
      {
        "type": "image",
        "name": "extra.png",
        "url": "https://..."
      }
    ],
    "type": "screenshot",
    "resource": "https://...",
    "status": "open",
    "membersOnly": false,
    "createdAt": "2026-07-10T14:22:00+00:00",
    "changedAt": "2026-07-10T15:00:00+00:00",
    "reporter": {
      "type": "reporter",
      "name": "Jane Client",
      "email": "jane@client.com"
    },
    "comments": [
      {
        "id": 7,
        "creator": {
          "type": "member",
          "name": "Alex",
          "email": "alex@agency.com",
          "id": 3
        },
        "body": "Looking into this.",
        "attachments": [],
        "membersOnly": false,
        "createdAt": "2026-07-10T15:00:00+00:00"
      }
    ],
    "sessionDetails": {
      "page": "https://example.com/about",
      "device": "mobile",
      "system": "iOS",
      "browser": "Safari",
      "screenWidth": 390,
      "screenHeight": 844,
      "viewportWidth": 390,
      "viewportHeight": 664,
      "devicePixelRatio": 3,
      "selector": "#logo"
    },
    "integrations": [
      {
        "slug": "jira",
        "url": "https://..."
      }
    ],
    "customMetadata": {},
    "console": {
      "error": 2,
      "warn": 1,
      "info": 0,
      "log": 5
    }
  }
}

Output details

changedAt is the latest of the feedback’s creation time, latest comment, or latest activity (title/tags/resolve/etc.).

Comments may also include a fromIntegration string when the comment originated from a synced integration.

Each comment includes an id. Pass that value as commentId to comment-update and comment-delete. attachments lists files on the comment. For files attached with attachment-create, each item has type (a MIME type such as image/png or text/plain), name (the original filename), and url (the file). An empty attachments array means the comment has no files.

Errors

• No team found for the authenticated user.
• Feedback not found.
• You do not have access to this feedback.

feedback-update

Tool name: feedback-update

What it does: Update a feedback item's title, description, tags, visibility, and/or status. Use status to resolve or reopen feedback. Only provided fields are changed. At least one of title, description, tags, membersOnly, or status must be provided.

Input schema

Output

{
  "feedback": {
    "id": 42,
    "title": "Logo is blurry on mobile",
    "description": "The logo looks pixelated on iPhone.",
    "tags": ["Bug"],
    "status": "resolved",
    "membersOnly": false
  }
}

Errors

• No team found for the authenticated user.
• Feedback not found.
• You do not have access to this feedback.
• At least one of title, description, tags, membersOnly, or status must be provided.
• Invalid tag(s): {tags}. Tags must exist on the project. Use the ProjectGet tool with projectId {id} to see available tags. Case sensitive.

Attaching files to a comment

Comment tools do not accept file bytes. Upload the file first, then pass the returned key when creating or updating the comment.

1. Call attachment-create once per file, with the original filename (including the extension).
2. PUT the file to the returned uploadUrl. Use the returned curl command and replace /path/to/local/file with the local path. The upload URL expires after 5 minutes. If it expires before the PUT succeeds, call attachment-create again and use the new URL and key.
3. Pass every returned key in the attachmentKeys array on comment-create or comment-update.

attachmentKeys values are UUIDs. Each key can be used once, and only by the same user on the same current team that created it. If a key is unknown, already used, expired, or belongs to someone else, the tool returns Invalid attachment key. and the comment is not created or changed.

There is no way to remove or replace files already on a comment. On comment-update, attachmentKeys only appends new files.

attachment-create

Tool name: attachment-create

What it does: Mint a short-lived upload URL for one comment attachment. PUT the file to uploadUrl, then pass key to comment-create or comment-update as attachmentKeys. This tool does not accept file bytes. Call it once per file. See Attaching files to a comment above.

Input schema

Output

{
  "key": "2e2cdb8b-9897-4a6c-a935-82922b1cfb87",
  "uploadUrl": "https://...",
  "curl": "curl -X PUT -T /path/to/local/file 'https://...'"
}

Output details

key is the value to pass in attachmentKeys after the file has been PUT to uploadUrl.

uploadUrl is a presigned URL. PUT the file bytes there. Do not POST them to Feedbucket. The URL expires after 5 minutes.

curl is ready to run. Replace /path/to/local/file with the local file path, then run it.

Errors

• No team found for the authenticated user.

comment-create

Tool name: comment-create

What it does: Create a new comment on a feedback item. If the feedback was resolved, commenting automatically reopens it. To attach files, follow Attaching files to a comment above, then pass the keys as attachmentKeys

Input schema

Output

{
  "comment": {
    "id": 8,
    "creator": {
      "type": "member",
      "name": "Alex",
      "email": "alex@agency.com",
      "id": 3
    },
    "body": "Fixed on staging — please recheck.",
    "attachments": [
      {
        "type": "image/png",
        "name": "shot.png",
        "url": "https://..."
      }
    ],
    "membersOnly": false,
    "createdAt": "2026-07-17T10:30:00+00:00"
  }
}

Output details

attachments is [] when the comment has no files. Each item has:

• type — MIME type from the filename extension, such as image/png or text/plain. Unknown extensions use application/octet-stream.
• name — the original filename.
• url — URL of the uploaded file.

When creator.type is member, creator also includes id.

Errors

• No team found for the authenticated user.
• Feedback not found.
• You do not have access to this feedback.
• Invalid attachment key.

comment-update

Tool name: comment-update

What it does: Update a comment on a feedback item. Only provided fields are changed. Get commentId from the comment id returned by feedback-get. attachmentKeys appends new files and does not replace existing ones. Editing a comment does not reopen resolved feedback. Any member of the feedback’s current team can update the comment, including a comment written by someone else. To attach files, follow Attaching files to a comment above.

Input schema

At least one of body, membersOnly, or attachmentKeys must be provided.

Output

Same comment object as comment-create. Existing files stay first in attachments. Newly appended files follow them.

{
  "comment": {
    "id": 8,
    "creator": {
      "type": "member",
      "name": "Alex",
      "email": "alex@agency.com",
      "id": 3
    },
    "body": "Updated: please recheck the logo.",
    "attachments": [
      {
        "type": "image/png",
        "name": "shot.png",
        "url": "https://..."
      },
      {
        "type": "text/plain",
        "name": "notes.txt",
        "url": "https://..."
      }
    ],
    "membersOnly": false,
    "createdAt": "2026-07-17T10:30:00+00:00"
  }
}

Output details

Errors

• No team found for the authenticated user.
• Comment not found.
• You do not have access to this comment.
• At least one of body, membersOnly, or attachmentKeys must be provided.
• Invalid attachment key.

comment-delete

Tool name: comment-delete

What it does: Delete a comment by ID. Get commentId from the comment id returned by feedback-get. Any member of the feedback’s current team can delete the comment, including a comment written by someone else. This does not delete the feedback item.

Input schema

Output

{
  "success": true
}

Errors

• No team found for the authenticated user.
• Comment not found.
• You do not have access to this comment.

feedback-get-console-log

Tool name: feedback-get-console-log

What it does: Get browser console log entries captured with a feedback item. Read-only and idempotent. Requires a plan that includes console logs.

Input schema

Output

{
  "consoleLog": [
    {
      "type": "error",
      "arguments": ["Uncaught TypeError: Cannot read property 'x' of undefined"],
      "captured_at": "2026-07-10T14:21:58+00:00"
    },
    {
      "type": "warn",
      "arguments": ["Deprecated API usage"],
      "captured_at": "2026-07-10T14:21:59+00:00"
    }
  ]
}

Output details

type is one of: log, warn, error, info.

Errors

• No team found for the authenticated user.
• Feedback not found.
• You do not have access to this feedback.
• Console log is not available on your plan.

How to Connect

Feedbucket MCP supports two authentication methods:

1. Personal access tokens — send as an Authorization: Bearer header. Used by Cursor and VS Code. See Access Tokens to create one manually.
2. OAuth — browser sign-in with dynamic client registration. Used by Claude Code, Claude Desktop, Claude.ai, ChatGPT, and Codex.

You can also follow the guided setup for each client in the dashboard at https://dashboard.feedbucket.app/account/mcp.
MCP endpoint used in all examples below:

https://dashboard.feedbucket.app/mcp

Cursor

Cursor connects with a personal access token.

Add to Cursor (one-click): On the MCP settings page, the Add to Cursor button creates an access token labeled "Cursor" and opens a deeplink (cursor://anysphere.cursor-deeplink/mcp/install?...) that installs the Feedbucket MCP server into Cursor with the token already configured.

Manual setup:

1. Create an access token (see Access Tokens).
2. In Cursor, open Settings → Tools & MCPs → New MCP Server.
3. Paste the following into mcp.json, replacing YOUR_TOKEN with your token:

{
  "mcpServers": {
    "feedbucket": {
      "url": "https://dashboard.feedbucket.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

VS Code

VS Code connects with a personal access token.

Add to VS Code (one-click): On the MCP settings page, the Add to VS Code button creates an access token labeled "VS Code" and opens a deeplink (vscode:mcp/install?...) that installs the Feedbucket MCP server into VS Code with the token already configured.

Manual setup:

1. Create an access token (see Access Tokens).
2. In VS Code, open the Command Palette and run MCP: Add Server, then choose HTTP — or paste the following into mcp.json, replacing YOUR_TOKEN with your token:

{
  "servers": {
    "feedbucket": {
      "type": "http",
      "url": "https://dashboard.feedbucket.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Claude Code

Claude Code connects with OAuth by default.
CLI (recommended):

claude mcp add --transport http feedbucket https://dashboard.feedbucket.app/mcp
claude mcp login feedbucket

The login command opens a browser (or prints a URL) for Feedbucket sign-in and authorization. If the redirect fails, paste the callback URL when prompted.
Token alternative: You can also use a personal access token via .mcp.json:

{
  "mcpServers": {
    "feedbucket": {
      "type": "http",
      "url": "https://dashboard.feedbucket.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Claude Desktop

Claude Desktop connects with OAuth.

1. Open Settings → Connectors.
2. Choose Add Custom Connector.
3. Name: Feedbucket.
4. URL: https://dashboard.feedbucket.app/mcp.
5. Add the connector and complete Feedbucket authorization in the browser.

Connectors are tied to your Claude account. If you already connected Feedbucket on Claude.ai, it will appear in Claude Desktop as well.

Claude.ai

Claude.ai connects with OAuth, using the same Custom Connector flow as Claude Desktop.

1. Open Settings → Connectors.
2. Choose Add Custom Connector.
3. Name: Feedbucket.
4. URL: https://dashboard.feedbucket.app/mcp.
5. Add the connector and complete Feedbucket authorization in the browser.

ChatGPT

ChatGPT connects with OAuth.

1. Open Settings → Apps → Advanced settings and enable Developer mode.
2. Go to Settings → Apps → Create.
3. Name: Feedbucket.
4. URL: https://dashboard.feedbucket.app/mcp.
5. Auth: OAuth.
6. Create the app and complete Feedbucket authorization.
7. In a new chat, enable Feedbucket from the + menu.

Codex

Codex connects with OAuth via the CLI.

codex mcp add feedbucket --url https://dashboard.feedbucket.app/mcp
codex mcp login feedbucket

The login command opens a browser (or prints a URL) for Feedbucket sign-in and authorization.

Access Tokens

Personal access tokens let clients authenticate to the MCP server with a Bearer header. Cursor and VS Code use this method (and their Add buttons create a token for you automatically). Claude Code can also use a token as an alternative to OAuth.

Create a token

1. Go to https://dashboard.feedbucket.app/account/access-tokens.
2. Create a token with a label (for example, Cursor or Local testing).
3. Copy the token immediately — it is shown only once and cannot be retrieved later.

Tokens are scoped to mcp:use and expire after one year.

Use a token

Send the token as an Authorization header on requests to the MCP endpoint:

Authorization: Bearer YOUR_TOKEN

Example MCP config (Cursor-style):

{
  "mcpServers": {
    "feedbucket": {
      "url": "https://dashboard.feedbucket.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Requests without a valid token receive 401 Unauthorized. Tokens that are missing the mcp:use scope receive 403 Forbidden.

Connected apps

The same Access Tokens page also lists applications you have connected via MCP using OAuth, so you can see and manage both token-based and OAuth clients in one place.

Updated on: 05/10/2026