Reviewers pin it. Your agent fixes it.
Send one link to a teammate, PM, designer or client. They click anything on your live site and leave a note, free, with no account. Your coding agent picks up each pin you approve through MCP and makes the fix.
Pincushion AI drops the first pins on your live page. Then invite your team or client, or hand the pins to your agent. No card, no extension. See a sample crit
Pincushion AI pins your live page, then your reviewers add theirs.
Anyone drops a pin
Your reviewer — teammate, PM, designer or client — clicks any element on your live app and types a note. No account, no paid seat, nothing to install. We capture the selector and screenshot behind the scenes.
Discuss in place
Threaded comments appear right on the page. Context stays where it belongs — not in a Slack thread from last Tuesday.
Your agent picks it up
The pin arrives via MCP in Cursor, Claude Code, Codex, or Windsurf — selector, screenshot, viewport, thread, and project context in one call. No translating screenshots into prompts.
The loop closes
Resolving records the commit, branch, and PR. A configured deploy hook links the live deploy and can queue eligible pages for a connected agent to check. Your reviewer sees the recorded verification result once that check completes.
Also works for native iOS apps: generate a mobile crit wall from simulator screens and reviewers comment in a browser. Or skip the browser entirely — a note or screenshot emailed to your project's pin+…@in.pincushion.io address becomes a pin.
The feedback loop,
finally closed.
An agent work packet, not a feedback item
Each pin reaches your coding agent in one MCP call: selector, DOM snippet, screenshot, viewport, thread, project context, and acceptance criteria. Resolved pins keep the commit SHA, branch, and PR URL attached to the implementation record.
Share a report, no install
Share a report at pincushion.io/r/<token>. It's a live screenshot of the page with real pin markers, threads, and status. Anyone with a public link can view it. Visitors can add pins and replies without an account only on projects open to feedback, after entering a name (email is optional). Restricted feedback and viewing private reports require authorized sign-in. No extension is needed. Click into a resolved pin to see what the page looked like before, the PR that fixed it, and when it shipped.
Pincushion AI
AI feedback on every plan. Run /critique and Pincushion AI drops UI/copy/a11y pins on your page. Mention @pincushion in any thread for a follow-up. Workspace critiques use your connected agent and its model provider. Hosted Crit sends selected rendered-page evidence to OpenAI for analysis. Resulting feedback syncs to Pincushion.
Works in the editor
you already use.
More ways to pin: the Chrome extension, Slack and Teams alerts, iOS simulator crit walls, and the REST API and n8n.
Your agent doesn't
just read feedback.
Pincushion exposes an MCP server your supported coding agent can connect to. Approve a finding, then ask the agent to read the pin, make the change, and record the result. Unattended dispatch requires a separately configured agent loop.
Get Started// Connect in one command npx pincushion-mcp --project-dir . // Agent fetches pins waiting on it get_actionable_pins() { "pins": [{ "selector": ".hero > h1", "comment": "Change to Invoicing for freelancers" }] } // Agent writes the change & resolves fix_and_resolve("pin_1", "Updated heading text") // done
See it on your own page first.
Not ready to wire up your agent? Drop your web URL and Pincushion AI pins concrete UI, copy, and a11y fixes right on your live page. Shipping native mobile? Generate a crit wall from simulator screens and use the same pin-to-agent loop. Keep the pins with a free account and invite your team — nothing to install on anyone's end.
Get my free crit →Free account, no credit card. Most crits take 2–3 minutes.
Connect your agent.
Run npx pincushion-mcp setup in your repo. It signs you in, links your site and sets up the editor you choose. Then ask your agent to read your pins.
Setup details
Setup signs in, links your site to a project and prepares the client you explicitly choose. With Node/npm, a supported client runtime and model access available, it continues into a real critique and verifies the saved report before opening its link. Normal workspace/MCP trust, cloud-upload and report-visibility prompts still apply; private reports may require browser sign-in. Missing access leaves the run incomplete. Manual configuration below alone creates no report. If this came from a Crit report, use npx pincushion-mcp claim <report-token> to resume after authenticated project access is confirmed. The Chrome extension is optional for adding new browser pins.
VS Code & Cursor extension
1-clickThe extension adds @pincushion as a Copilot Chat participant. The MCP button adds the Pincushion server to VS Code in one click. Run the setup wizard separately to sign in, register the project, and choose an editor; restart the editor and read a real pin before treating the agent as connected.
No Marketplace access? Download the VSIX (v0.2.4) and run code --install-extension pincushion.pincushion-feedback-0.2.4.vsix, or drag it into Extensions → Install from VSIX….
@pincushion /pins — show all open feedback@pincushion /my-pins @you — your @mentioned pins@pincushion /resolve <id> — claim & close a pin@pincushion /feedback-summary — project overview@pincushion /implement — run the full approved-pin loop
After a configured deployment, the deploy webhook can queue eligible pages for critique. A connected agent processes the queue and records the verification result. Queueing depends on project configuration, license status, available quota and capture candidates.
Cursor
1-clickChoose Cursor in the setup wizard, or add the MCP server with this explicit one-click action.
Add Pincushion MCP to CursorOr add manually to the project-local .cursor/mcp.json:
Claude Code
PluginInstall the Pincushion plugin from its marketplace. It adds the MCP server and the setup, pins, implement and resolve skills.
Only want the MCP server? Run this in your project root instead.
Restart Claude Code, then ask it to call get_actionable_pins. Review an existing AI or human finding, approve that exact pin, and then ask it to implement the change.
Claude Desktop
Config fileAdd to claude_desktop_config.json. On macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Replace /absolute/path/to/project with your repo path. Restart Claude Desktop after saving.
Windsurf
Config fileChoose Manual in the setup wizard, then open Windsurf → Preferences → MCP, or edit ~/.windsurf/mcp.json directly. The wizard leaves unsupported editor configs unchanged.

Antigravity
Config fileChoose Manual in the setup wizard, then add this to your Antigravity MCP config file. Use . for the project dir if running from your repo root, or provide an absolute path.
Any MCP Client
npxAny editor or agent runtime with MCP support can connect using npx. Choose Manual in the wizard and follow that client’s MCP instructions; point --project-dir at your repo root.
Also works with:
+ Zed, Cline, Continue, Goose, Amp, Codex CLI, and any MCP client
Full setup instructions on desktop.
Built for the fix,
not the ticket.
| Bug-report tools with MCP | pincushion | |
|---|---|---|
| First pins | Someone files a report | Pincushion AI pins your live page, then your reviewers add theirs |
| What your agent gets | Screenshot, URL and browser info, plus console logs (Marker.io) or the CSS selector (BugHerd) | Selector, DOM snippet, screenshot, viewport, thread and acceptance criteria in one MCP call |
| After the fix | The agent updates the issue (BugHerd) or opens a PR (Marker.io) | The pin keeps the commit, branch and PR, and resolves on deploy once the deploy hook is set up |
| Proof it shipped | Not in their MCP docs | After a configured deploy, a connected agent records verified, regressed or inconclusive on the pin (every plan; unlimited on Pro and up) |
Compared with Marker.io's and BugHerd's own MCP docs, October 2026. Both are solid bug reporters. Pincushion is built for the step after the report.
Independent contributors pinned it.
OpenClaw merged the fix.
“It would be nice if there was option to easily copy this command, so users didnt have to type it out.”
- Pincushion AI and independent contributors to the open-source OpenClaw project pinned its iOS onboarding and chat screens.
- Pincushion’s founder opened PR #99243 with fixes from the approved pins, including a copy button for that command.
- An OpenClaw maintainer merged it on July 4, 2026. The PR cites the crit as its baseline and kept only the fixes that held up.
Start free.
Scale when ready.
Stakeholders are always free: unlimited reviewers, no seats, nothing to install. You only pay to unlock more projects and Realtime. No hidden fees, cancel anytime.
Free
Stakeholders always free. Devs get the full agent-native workflow on one project.
- Unlimited stakeholder seats — free forever
- Unlimited manual pins, comments & the full MCP workflow
- 25 AI actions / month — agent dispatches, AI replies, AI critique pins & deploy verifications (manual work never counts)
- Agent work packets — pins carry selector, DOM, viewport, thread & project context
- Assign-to-agent — one MCP call dispatches the pin to your local agent loop
- Branch and review link recorded on resolved pins
- Implementation packets (pins grouped by page)
- Slack & Teams routing for ready pins & mentions
- 1 project · Cloud sync (30s polling) · Slash commands for all IDEs
Pro
For teams shipping with AI coding agents
- Unlimited AI actions — no monthly pool on dispatches, AI replies, critiques & verifications
- Realtime push to your coding agent — pin → IDE in under a second
- Auto-link deploy URL — every deploy writes the live URL back to its resolved pins
- Post-deploy verification — Your connected agent can check eligible deployed fixes and write a verdict (
verified,regressed, orinconclusive) on the pin - Time-to-fix metrics for your projects — median + p25/p75 of pin-to-resolve, with a 5-pin minimum so it's never noise
- Access controls — open, domain-locked, or invite-only
- Unlimited projects
- 1 developer seat, plus unlimited free commenters
- Everything in Free (agent work packets, assign-to-agent, branch and review links, AI critic, Slack/Teams)
- Priority email support
Upgrade anytime; paid plans bill right away
Team
For product teams with a few developers
- 3 developer seats included
- Additional developer seats at $12/mo each ($10/mo billed annually)
- Unlimited commenters — free forever
- Everything in Pro (Realtime, auto-link deploy, post-deploy verification, time-to-fix metrics, access controls)
- Role management (owner, editor, commenter)
- @mention routing between team members
- Priority support
Upgrade anytime; paid plans bill right away
For agencies
Agency
For studios and agencies running client work across many sites
Upgrade anytime; paid plans bill right away
Already on Team? You keep white-label reports. New Team plans don’t include them.
- Everything in Team
- 5 developer seats included, then $12/mo each ($10/mo billed annually)
- Unlimited client projects
- White-label crit reports — your logo and studio name on client-facing reports
- Internal threads & private labels — notes your clients never see
- Evidence export — CSV and PDF of pins, reviews and verification
- Client-review activity — who approved which review version
- Priority support
See full pricing, extra seats and FAQ →
Stop screenshotting your product into Slack.
Just pin it.
Visual comments, dropped on the live thing — and shipped by your agent.
Get my free crit