Your AI Agents can now act on comments added in BugSmash with our MCP. See How it works.
Model Context Protocol · OAuth 2.1

Your reviews,
inside your AI client

Connect Claude, Cursor, Windsurf or ChatGPT to BugSmash. Ask for open feedback, turn comments into assigned tasks and push approvals forward without switching tabs.

Works with
Claude Cursor Codex ChatGPT Windsurf Cline Zed
Claude · bugsmash connected
Summarise the open feedback on the Acme homepage redesign.
get_feedback · project: homepage-redesign · 14 comments
14 open comments. 9 are design: hero headline spacing, button contrast on mobile, inconsistent card radii, misaligned pricing grid. 3 are copy, 2 are bugs.
Fix all the design ones and raise a PR.
create_task · 9 design items grouped
update_status · 9 moved to in progress
Pushed branch fix/design-feedback with all 9 changes and opened PR #482. Each commit links back to its BugSmash comment, and the reviewers were notified for sign-off.

What teams get when BugSmash lives in their AI client

The MCP server removes the manual middle of the review loop, so feedback turns into shipped work in the same conversation.

9x

Faster feedback triage

Your assistant reads every annotated comment and groups it by theme in seconds instead of a manual sweep.

0

Context switches

Review, fix and ship from the same chat window. No tab hopping between tool, tracker and editor.

70%

Less manual ticket writing

Comments become properly titled, assigned and prioritised tasks without anyone retyping them.

1 prompt

From feedback to pull request

Ask once and the model fixes the design notes, links each commit to its comment and opens the PR.

Connected in under a minute

Use the same lightweight setup pattern your team already expects from MCP clients. Add the server, approve access and start asking.

  1. 1Add the serverPaste the BugSmash MCP URL into your client config.
  2. 2Approve OAuthSign in once and grant the workspace scopes you want.
  3. 3Start askingYour assistant can now read feedback and act on it.
claude_desktop_config.json
{
  "mcpServers": {
    "bugsmash": {
      "url": "https://api.bugsmash.io/mcp"
    }
  }
}
OAuth ready 6 tools enabled

All the Tools for full review loop

Assistants can browse the structure of your review workspace, read the exact feedback thread, then create the next project, version, comment or reply when your workspace allows writes.

Read tools keep the assistant grounded in folders, projects, versions and discussions before it acts.

Write tools stay explicit and permissioned, covering creation plus resolution and approval status changes.

  • list_foldersreadBrowse folders.
  • list_projectsreadView project status and ownership.
  • list_versionsreadSee uploaded revisions.
  • list_commentsreadPull pinned feedback context.
  • list_repliesreadRead comment discussions.
  • create_replywriteReply in a feedback thread.
  • create_commentwriteAdd a new review comment.
  • update_commentwriteChange resolution status.
  • create_folderwriteCreate an organizing folder.
  • create_projectwriteStart a new review project.
  • create_versionwriteAdd a new revision.
  • update_projectwriteChange approval status.

Chat that actually moves the project forward

MCP turns your AI client into a live review workspace: it can read the source context, act with permission and keep the work moving.

Read and write feedback

Assistants list open comments, reply in thread and add new annotated notes without leaving chat.

Drive approval workflows

Change review status, request sign-off and trigger hand-off steps from a single instruction.

Always current context

Version history, activity and task state are fetched live, so the model never works from stale data.

Scoped by permission

OAuth 2.1 tokens mean an assistant can only reach what the connected user can already see.

Works with the clients your team already uses

Standard MCP over HTTP and SSE means every new compatible client works on day one, with no plugin to install and nothing to self-host.

  • No self-hosting, no local proxy
  • Per-workspace scopes and revocable tokens
  • Works with any MCP-compatible client
Claude
Cursor
Codex
ChatGPT
Windsurf
Cline
Zed

Frequently asked questions

What is the BugSmash MCP server?

It is a Model Context Protocol server that exposes your BugSmash projects, feedback and approvals as callable tools for any compatible AI assistant.

Get your unbilled hours back

Start today, it's free

Set up your first review in under a minute. No credit card, no onboarding calls.

BugSmash website review interface with comment controls