ReviseMy Try with your agent

Connectors

Connect ReviseMy to the agent you already use

One MCP endpoint and a Bearer try token. Paste the config into ChatGPT, Claude, Copilot, Cursor, or Grok — then run a design checkup without leaving your workflow.

The problem

Agent setup docs often live on GitHub while the product lives on a marketing homepage. You need a crawlable place that explains which hosts support inline review, which share a link, and how to plug in without guessing.

How ReviseMy fits

  1. 1

    Get a free try token on the homepage and copy the host-specific MCP config.

  2. 2

    create_review opens inline on MCP Apps hosts, or shares a review_url on CLI and link hosts.

  3. 3

    get_review polls until you decide; next_action tells the agent what to do next.

Hosts

Same MCP endpoint for every host. Generate a try token on the homepage to copy ready-made config.

ChatGPT

Remote MCP or REST

Add ReviseMy as a remote MCP connector (or Custom GPT Action for REST). Use the MCP URL and Bearer token from your try token — ChatGPT’s UI takes those fields, not a full JSON paste. When the host supports MCP Apps, the review can render inline; otherwise the agent shares a `review_url` link.

  1. In ChatGPT, open Settings → Connectors (or Custom GPT → Actions for REST)
  2. Add a remote MCP connector named revisemy
  3. Paste the MCP URL and Authorization: Bearer {try_token}
  4. Paste a starter prompt asking it to run create_review and follow next_action

Starter prompt

Run a ReviseMy design checkup on the work I just changed. Use create_review with the right source (screenshots, public URL + capture_url, email HTML, or PDF), share the review_url if you get one, wait for my marks, then poll get_review and follow next_action until I approve.

Claude

Inline (Desktop) · Link (Code)

Claude Desktop can render the review inline via MCP Apps. Use Settings → Developer → Edit Config with the mcp-remote JSON (Bearer try tokens do not fit Connectors → Add custom connector, which is OAuth-oriented). Claude Code is CLI-only and shares a `review_url` instead.

  1. Desktop: Settings → Developer → Edit Config, merge the mcp-remote JSON, quit and reopen Claude
  2. Claude Code: run `claude mcp add --transport http …` with your try token
  3. Paste a starter prompt (or use design_checkup_loop) and mark / approve on Desktop inline or via review_url on Code

Starter prompt

Run a ReviseMy design checkup on the work I just changed. Call create_review with the right source. Open the review inline if available, otherwise give me the review_url. Wait for my marks, then follow next_action until I approve. Prefer the design_checkup_loop prompt if available.

Copilot

Inline MCP Apps

Paste or merge the servers JSON into Copilot MCP settings (user or workspace mcp.json). After create_review, the review renders inline in Copilot Chat so you can mark and approve without leaving the editor.

  1. Open Copilot → MCP
  2. Paste the try-token Copilot config from the homepage under servers
  3. Paste a starter prompt asking Copilot to call create_review and open the review inline

Starter prompt

Run a ReviseMy design checkup on the work I just changed. Call create_review with the right source (screenshots, public URL + capture_url, email HTML, or PDF). Open the review inline so I can mark and approve, then follow next_action until I’m done. Prefer the design_checkup_loop prompt if available.

Cursor

Link via review_url

Cursor agents use MCP tools in the IDE. After create_review, the agent shares a review_url — open it in the browser to mark feedback and approve. No inline MCP Apps UI in Cursor yet; the loop is the same.

  1. Open Cursor Settings → MCP
  2. Paste mcpServers JSON (or merge into ~/.cursor/mcp.json)
  3. Paste a starter prompt asking the agent to call create_review and share the review_url

Starter prompt

Run a ReviseMy design checkup on the work I just changed. Call create_review with the right source (screenshots as data URLs for localhost, or public URL + capture_url). Give me the review_url to mark and approve, then poll get_review and follow next_action until I approve. Prefer the design_checkup_loop prompt if available.

Grok

Custom connector · Link

Add ReviseMy as a custom MCP connector at grok.com/connectors. Paste the MCP URL and Bearer Authorization header (not the full JSON). After create_review, open the review_url to mark and approve (link workflow).

  1. Go to grok.com/connectors → New Connector → Custom
  2. Paste the public MCP URL and authorize with the Bearer token
  3. Paste a starter prompt asking Grok to call create_review and share the review_url

Starter prompt

Run a ReviseMy design checkup on the work I just changed. Call create_review with the right source, give me the review_url to mark and approve, then poll get_review and follow next_action until I approve.

Get a try token and copy config on the homepage — then come back here if you need the host overview.

What you get once connected

MCP Apps vs link hosts

Claude Desktop, claude.ai, and Copilot can render the review inline. Cursor, Claude Code, and Grok use the review_url tab — same marks, board, and next_action.

One try token, any host

Homepage try tokens mint a Sanctum Bearer for /mcp/revisemy and the REST API. Host packaging is install UX on top of the same Cloud-hosted endpoint.

Webhooks for CI/CD

Pass webhook_url to create_review. ReviseMy POSTs a signed review.decided event when you approve or request changes — gate pipelines instead of polling.

REST when MCP is not an option

Custom GPT Actions and scripts can use POST /api/reviews with the same Bearer token. Tool names and payloads stay host-agnostic.

FAQ

Do I need a ReviseMy account?

No. Grab a free try token on the homepage, paste the MCP config, and go. Reviewers only need the secret /r/{token} link.

Where do I get the config JSON?

On the homepage Try with your agent section. Generate a try token, pick your host tab, and copy the ready-made MCP URL, Bearer token, or JSON.

Can any MCP client connect?

Yes. Any client that speaks HTTP MCP can use /mcp/revisemy with Authorization: Bearer {try_token}. Host-specific tabs are shortcuts, not a closed list.

Try with your agent