Use the slicer MCP server
The MCP server is available to every slicer account. You sign in with yours the first time you connect.
Connect your AI agent to Competitor Watch using the Model Context Protocol (MCP), an open standard that lets agents call outside tools. Your agent can read what changed on competitor funnels, look at the screenshots, ads and emails behind those changes, manage the products you watch and start new runs.
The server is at:
https://mcp.slicer.dev/mcpSet up
Every agent signs in with your slicer account, and slicer asks you to approve the connection once.
Claude
For claude.ai, Claude Desktop, Cowork and mobile.
- Open the Add custom connector dialog. It opens with slicer's name and URL filled in.
- Check the details and add the connector.
- Choose Connect and sign in to slicer.
On a Team or Enterprise plan, an Owner adds the connector first. Each member then connects with their own slicer account.
Claude Code
For the terminal, VS Code and JetBrains. Run this in your terminal:
claude mcp add --transport http slicer https://mcp.slicer.dev/mcpThen run /mcp in Claude Code and sign in to slicer.
Codex
For the CLI, IDE extension and desktop app. Add the server:
codex mcp add slicer --url https://mcp.slicer.dev/mcpThen sign in to slicer:
codex mcp login slicerChatGPT
Your ChatGPT plan or workspace settings can turn custom servers off.
- In ChatGPT, open Plugins, select the plus button and choose Add custom MCP server.
- Name it
slicerand paste the server URL:https://mcp.slicer.dev/mcp - Choose OAuth if it asks how to sign in, then create it as a plugin.
- Sign in to slicer when ChatGPT asks.
OpenCode
Add this to opencode.json:
{
"mcp": {
"slicer": {
"type": "remote",
"url": "https://mcp.slicer.dev/mcp",
"oauth": { "callback_port": 19876 }
}
}
}Then sign in to slicer:
opencode mcp auth slicerOpenCode signs in through local port 19876, so that port has to be free when you sign in.
Things to ask
Once connected, try asking your agent:
- What did my competitors change on their funnels this week? Show me the screens behind the biggest changes.
- Start watching https://example.com/quiz, then tell me what its funnel looks like once the first run has finished.
- Which of my competitors' ads have been running longest, and which funnels do they link to?
- Read the follow-up emails from my competitors and tell me who offers a discount, and how far into the sequence.
- Find competitors of example.com and add the ones that sell the same thing.
How it works
- Your agent acts as you. It sees the products on your account and nobody else's, and it can't see or change your billing or password.
- Runs take a while.
add_product,run_productandrun_groupstart the work and return at once, and your agent checks back later withget_runorlist_products. - A competitor search can take up to 15 minutes, and only one runs at a time.
- Only "No changes" means nothing changed. A funnel that couldn't be compared says why.
- Text and images from competitor pages reach your agent labelled as data, so it doesn't follow instructions written on a page.
Tools
Reading data
| Tool | Description |
|---|---|
list_products | Your groups and products, with each product's latest run and next check. |
get_run | A run's status and how many funnels it has captured so far. |
get_changes | What changed on competitor funnels in a week, with dates and source URLs. |
get_funnel | One funnel from one capture: the analysis, then every screen as text. |
get_screenshots | Up to 3 screens of a funnel as images. |
get_ads | The Meta ads found in the latest check: copy, format, how long each has run and where it links. |
get_ad_image | One ad's creative, or a video's preview frame. |
get_emails | The follow-up emails competitors sent after the latest check signed up. |
get_email_image | One email as it was received. |
get_product_search | Whether a competitor search is still going, and what it has found. |
Changing data
| Tool | Description |
|---|---|
add_product | Start watching a competitor from its funnel or landing URL, with a first run straight away. |
add_found_products | Start watching products a finished search found. |
remove_product | Stop watching a product. Adding it again later starts a new history. |
run_product | Start a run for one product. |
run_group | Start runs for every idle product in a group. |
find_products | Search for competitors of a product from its website or landing page. |
Limits
Each account has these limits. When your agent hits one, it gets a message saying when the limit resets.
| Action | Limit |
|---|---|
| Products added | 20 a day |
| Product runs started | 30 a day |
| Competitor searches | 20 a day, shared with the web app |
| Reads | 300 a minute |
Troubleshooting
The sign-in page says "This app can't connect"
That app isn't one of the agents listed on this page. Connect from one of those instead.
The sign-in page says "This request has expired"
Go back to your agent and connect again.
My agent asks me to sign in again
Repeat its sign-in step: /mcp in Claude Code, codex mcp login slicer in Codex, opencode mcp auth slicer in OpenCode, or Connect on the connector in Claude.
My agent says there is nothing to read yet
A product has no results until its first run finishes. Ask your agent to check it with list_products, and to start one with run_product if it has none.
A run says it is queued
slicer is full right now. The run starts on its own when a slot frees up, so your agent doesn't need to start it again.
My agent says I reached a limit
The message says when that limit resets. Retrying before then won't help.
Something else
Email hello@slicer.dev with the agent you use and what it told you.
Disconnect
Open Settings, find the agent under Agents and revoke it. That ends its sign-in, and a token it already holds keeps working for up to an hour.
Privacy
What your agent reads goes to the company that runs your agent, under that company's terms. slicer sees the tool calls your agent makes and never your conversation with it. We record which tool was called, whether it worked and which agent called it. Our privacy policy has the details.
Last updated November 7, 2026