Model Context Protocol

Your site’s growth data,
in your AI assistant.

Mimetic connects AI assistants to website audits, analytics, advertising, and experiment results to identify revenue opportunities and implement reviewable improvements.

Connect the accounts you use, then ask about traffic, revenue, search, ad performance, visitor friction, and A/B test results. Tools operate on sites your Mimetic account can access.

Connect your assistant · Download the tool catalog (JSON)

Connect to Mimetic

Remote MCP

Add this URL in a client that supports Streamable HTTP and OAuth:

https://mcp.trymimetic.com/mcp

Your client discovers OAuth 2.1, opens Mimetic sign-in, and asks you to approve access. It uses PKCE (S256); access tokens last an hour and can be refreshed. You can revoke access from your account.

CLI and local MCP clients

npm install -g @mimeticinc/mim-cli
mim setup

Setup signs you in and registers MCP with supported installed clients: Claude Code, Codex, Cursor, Windsurf, and Claude Desktop. See the client setup guide and npm package.

Start with list_projects, choose your site, then call get_growth_overview. Connections and plan requirements are described by each tool. See pricing.

What to ask

Find revenue opportunities

Review my site’s audit findings and connected analytics. Which opportunities should we investigate first?

Tools: list_projects, get_growth_overview, query_ga4

Investigate landing-page performance

Compare landing-page revenue and sessions over the last 28 days with the previous period.

Tools: query_ga4, query_gsc

Use experiment results

Review results from my connected experimentation platform and suggest what to test next.

Tools: query_experiments

Review visitor friction

Review recent recordings for recurring bugs and UX friction, with evidence for each proposed change.

Tools: review_recent_recordings

Set up missing measurement

Check what tracking my site needs and explain the setup changes before making them.

Tools: get_connection_recommendations, setup_analytics, analytics_setup_status

Example tool request

After authenticating, this request asks for your connected site’s landing-page sessions and revenue. Replace your-site.com with a project returned by list_projects.

{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"query_ga4","arguments":{"project":"your-site.com","metrics":["sessions","totalRevenue"],"dimensions":["landingPage"],"start_date":"28daysAgo","end_date":"yesterday"}}}

The response contains your requested GA4 report rows. Audit findings require verification, recording insights cover a consent-gated sample, and experiment conclusions depend on the connected provider’s measurements.

Access and permissions

mcp:read allows project data and reports. mcp:write is required for setup and configuration actions. A connected provider may require its own consent. Tool calls enforce account, project, and plan access.

Use the descriptions below to understand each action before approving it. Code fixes use the Mimetic CLI or report workflow and produce work you can review. MCP returns recording summaries and redacted insights; it does not return raw replay event payloads.

OAuth resource metadata · Authorization server metadata · Privacy

Live tool catalog

These 41 definitions come from the same catalog as MCP tools/list. This page and its JSON version are public. Connecting to MCP and executing tools require authentication.

List Projects
list_projects
Read · mcp:read
List TryMimetic projects/sites available to this token.
Input schema
```
{
  "additionalProperties": false,
  "properties": {},
  "type": "object"
}
```
Growth Overview
get_growth_overview
Read · mcp:read
Return audit findings, recent recording summaries, and agent guidance for a project. This is a COMPOSITE of other tools and inherits their limits: findings are heuristic detections rather than verified defects, and recordings are a consent-gated and therefore biased sample. Lead with the growth or revenue outcome and keep the technical detail underneath as evidence.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "maximum": 100,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}
```
List Audit Findings
list_audit_findings
Read · mcp:read
List audit/backlog findings for a project. Findings are DETECTIONS, not verified defects, and severity is scored by heuristic. Do not report one as a confirmed bug without checking the live page: individual detectors have known false-positive classes. The audit sees only what is publicly observable from outside, so a clean result is a lower bound and the ABSENCE of a finding is not evidence the problem is absent. Distinguish broken from cosmetic from subjective before recommending work, and describe fixes as proposals to test, never as ready to ship.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "maximum": 100,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}
```
List Session Recordings
list_recordings
Read · mcp:read
List recent session recording summaries for a project. Recording is consent-gated, so sessions are a BIASED SAMPLE of traffic: never infer total volume, conversion rate, or 'most users' from them, and never read an absence of recordings as an absence of traffic. Counts here are small, so give the plain count rather than a percentage, and do not call a pattern seen in a handful of sessions a trend.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "maximum": 100,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}
```

Review Recent Recordings

review_recent_recordings
Read · mcp:read

Use when the user asks Claude Code to check the last N recordings, find bugs, identify UX friction, or propose A/B tests. Returns a code-review brief with recording evidence, patterns, audit context, and suggested output format.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "focus": {
      "enum": [
        "bugs",
        "ab_tests",
        "both"
      ],
      "type": "string"
    },
    "limit": {
      "maximum": 100,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}

Get Replay Insight
get_replay_insight
Read · mcp:read
Return the persisted replay insight for a session. The narrative is MODEL-GENERATED from the recording and redacted, so it is an interpretation and not a transcript: never quote it as something the user said or did verbatim, and never present a single session as representative. Redaction means fields can be missing by design, which is not the same as nothing having happened.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "latest": {
      "type": "boolean"
    },
    "project": {
      "type": "string"
    },
    "session_id": {
      "type": "string"
    }
  },
  "type": "object"
}
```

Query GA4

query_ga4
Read · mcp:read

Run a READ-ONLY Google Analytics 4 report for a project's connected property. Returns rows for the requested GA4 metrics/dimensions over a date range. Use to answer questions about the user's real traffic, conversions, revenue, channels, landing pages, devices, etc.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "dimensions": {
      "description": "GA4 dimension API names, e.g. date, sessionDefaultChannelGroup, pagePath, landingPage, country, deviceCategory.",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "end_date": {
      "description": "Default today.",
      "type": "string"
    },
    "limit": {
      "maximum": 1000,
      "minimum": 1,
      "type": "integer"
    },
    "metrics": {
      "description": "GA4 metric API names, e.g. sessions, activeUsers, screenPageViews, conversions, totalRevenue, engagementRate.",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "order_by_metric": {
      "description": "Sort rows by this metric, descending.",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "start_date": {
      "description": "GA4 date or NdaysAgo (e.g. 28daysAgo, 2024-01-01). Default 28daysAgo.",
      "type": "string"
    }
  },
  "required": [
    "project",
    "metrics"
  ],
  "type": "object"
}

Query Search Console

query_gsc
Read · mcp:read

Run a READ-ONLY Google Search Console query for a project's verified site. Returns rows with clicks, impressions, ctr, position for the requested dimensions. Use for SEO/search questions: top queries, pages, countries, devices.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "dimensions": {
      "description": "GSC dimensions: query, page, country, device, date. Default [query].",
      "items": {
        "type": "string"
      },
      "type": "array"
    },
    "end_date": {
      "description": "YYYY-MM-DD. Default ~3 days ago (GSC data lag).",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "row_limit": {
      "maximum": 1000,
      "minimum": 1,
      "type": "integer"
    },
    "start_date": {
      "description": "YYYY-MM-DD. Default ~90 days ago.",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Describe The BigQuery Export
describe_bq
Read · mcp:read
Describe the project's GA4 BigQuery export before writing SQL for query_bq: the dataset name (analytics_<property id>), its location, and the date range of the daily events_YYYYMMDD and intraday tables, plus example SQL and the query limits. The dataset comes from the project's selected GA4 property.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Query BigQuery

query_bq
Read · mcp:read

Run a READ-ONLY GoogleSQL SELECT against the project's own GA4 BigQuery export (raw events: event_params, user_pseudo_id, traffic_source, items, ecommerce). Unqualified tables resolve to the project's dataset, so FROM `events_*` with a _TABLE_SUFFIX filter works. SELECT only, the project's own dataset only (checked with a BigQuery dry run), capped bytes billed and rows. Returns columns and rows. Run describe_bq first for the dataset and available dates.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "limit": {
      "description": "Maximum rows returned (default 100).",
      "maximum": 1000,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "sql": {
      "description": "GoogleSQL SELECT, e.g. SELECT event_name, COUNT(*) AS n FROM `events_*` WHERE _TABLE_SUFFIX \u003e= \u002720260901\u0027 GROUP BY 1 ORDER BY n DESC",
      "type": "string"
    }
  },
  "required": [
    "project",
    "sql"
  ],
  "type": "object"
}

List GA4 Properties
list_ga4_properties
Read · mcp:read
List every GA4 property the project's connected Google account can see: id, name, account, website URL, sessions over the last 28 days (a dead or wrong property is obvious next to its live sibling), and which property the project currently reads from. Use before set_ga4_property when numbers look wrong or several properties claim the same site.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Switch GA4 Property

set_ga4_property
Action · mcp:write

Point the project's Google Analytics connection at a specific GA4 property. Validates the id against what the connected account can see, then persists it; every query_ga4 read and Mimetic dashboard uses the new property from then on. Run list_ga4_properties first and pick the property with real traffic.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "property_id": {
      "description": "Numeric GA4 property id from list_ga4_properties, e.g. 403226097.",
      "type": "string"
    }
  },
  "required": [
    "project",
    "property_id"
  ],
  "type": "object"
}

Set Up Analytics
setup_analytics
Action · mcp:write
One-click hosted analytics setup for a project's site, so the owner never touches a Google console: Mimetic provisions and hosts the GA4 property and grants the owner Viewer access to it, installs the tag via a GitHub PR or the Shopify app pixel, ships Search Console verification, and submits sitemaps to Google/Bing (plus IndexNow) once verified. Use when a site has no analytics or the user asks to set up GA4 / Search Console. The run takes a minute or two, so this starts it in the background: poll analytics_setup_status for progress. If YOU are a coding agent with the site's repo checked out, pass channel='agent': no PR pipeline or GitHub App needed, provisioning runs server-side and analytics_setup_status returns the exact snippet + placement instructions for you to apply and commit yourself.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "channel": {
      "description": "auto (default): install via GitHub PR / Shopify app. agent: you apply the tag in the repo yourself; fetch the snippet from analytics_setup_status once provisioning finishes.",
      "enum": [
        "auto",
        "agent"
      ],
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "type": "object"
}
```
Set Up Google Ads
setup_google_ads
Action · mcp:write
Stand up a Google Ads account for a project's site with the tracking already wired: created under Mimetic's manager account so there is nothing to configure, linked to the site's GA4 property, auto-tagging on, and an ADMIN invite emailed to the site owner. The owner holds admin on the account and the manager link stays in place until they remove it. Ads can run once the invited admin accepts and adds billing. Reuses an already-linked account instead of creating a duplicate. Requires a GA4 property, so run setup_analytics first if there is none. Runs in the background: poll google_ads_status.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "type": "object"
}
```
Google Ads Status
google_ads_status
Read · mcp:read
Whether a Google Ads account is linked for this project: customer ID, account name/currency when readable, which credentials can query it, and the most recent ads setup event (created/linked/failed). If nothing is linked, says how to set one up.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "type": "object"
}
```

Query Google Ads

query_google_ads
Read · mcp:read

Run a READ-ONLY Google Ads (GAQL) report against the project's linked Ads account: campaign spend, clicks, conversions, search terms, ad groups, etc. Accepts a full GAQL SELECT query; defaults to campaign performance over the last 30 days. cost_micros/*_micros fields are millionths of the account currency.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "gaql": {
      "description": "GAQL SELECT query, e.g. SELECT campaign.name, metrics.cost_micros FROM campaign WHERE segments.date DURING LAST_7_DAYS.",
      "type": "string"
    },
    "limit": {
      "maximum": 1000,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Analytics Setup Status
analytics_setup_status
Read · mcp:read
Status of the hosted analytics setup for a project: none, in_progress, done (installed, waiting on data), live (events flowing), failed, or already_connected, plus the GA4 measurement ID, install-PR review URL, and Search Console verification state. For channel='agent' setups, the payload includes agent_install with the exact snippet + placement instructions for you to apply in the repo. Checking status also retries any pending Search Console verification and sitemap-submission followups.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "type": "object"
}
```
Set Up Session Recording
setup_session_recording
Action · mcp:write
Install Mimetic's session recording on a project's site: a masked, consent-gated loader (records NOTHING until the site's consent tooling grants analytics consent; IAB TCF, Cookiebot, and OneTrust are pre-wired). channel='auto' ships it as a GitHub PR; channel='agent' returns the snippet + placement instructions for you to apply in the repo. Insights then flow to list_recordings / review_recent_recordings and the daily Slack digest.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "channel": {
      "description": "auto = GitHub PR; agent = you apply the returned snippet yourself.",
      "enum": [
        "auto",
        "agent"
      ],
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```
What To Connect Next
get_connection_recommendations
Read · mcp:read
What to connect or set up NEXT for a project, filtered to what the site's detected stack makes relevant: GA connect only when GA4 exists (hosted setup_analytics when nothing does), Google Ads only when linkable, connect_posthog only when PostHog is detected, and a Klaviyo partner signup link only when NO email/SMS tool is detected. Call this before suggesting connections so users aren't overwhelmed with irrelevant asks.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Connect A/B Testing Platform

connect_experimentation
Action · mcp:write

Connect a project's A/B-testing platform (Intelligems, VWO, Optimizely, or GrowthBook). Optimizely and VWO connect over OAuth when configured: the tool returns an authorization URL for the user to open and approve, like connect_klaviyo. Intelligems and GrowthBook have no OAuth — call once with just vendor to get API-key instructions, then again with api_key. Credentials are validated against the live vendor API, stored encrypted, and bound to your token's email. Read-only usage only. VWO needs account_id (numeric, in the app.vwo.com URL); Optimizely takes account_id (project id) when the token sees multiple projects; self-hosted GrowthBook takes api_host.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "account_id": {
      "description": "VWO account id / Optimizely project id.",
      "type": "string"
    },
    "api_host": {
      "description": "Self-hosted GrowthBook API host, e.g. https://gb-api.example.com.",
      "type": "string"
    },
    "api_key": {
      "description": "Vendor API key/token \u2014 only for the key-paste path.",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "vendor": {
      "enum": [
        "intelligems",
        "vwo",
        "optimizely",
        "growthbook"
      ],
      "type": "string"
    }
  },
  "required": [
    "project",
    "vendor"
  ],
  "type": "object"
}

Query Experiments

query_experiments
Read · mcp:read

READ-ONLY reads from the project's connected A/B-testing platform (connect_experimentation first). action=list returns a normalized experiment/campaign list (id, name, status); action=results returns the vendor's results payload for one experiment_id (lift, conversions, revenue-per-visitor for Intelligems; variation reach and metrics for Optimizely; phase results for GrowthBook; campaign detail for VWO). If vendor is omitted and the project has exactly one connected platform, that one is used.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "Default list.",
      "enum": [
        "list",
        "results"
      ],
      "type": "string"
    },
    "experiment_id": {
      "description": "Required for action=results (from action=list).",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "vendor": {
      "enum": [
        "intelligems",
        "vwo",
        "optimizely",
        "growthbook"
      ],
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Disconnect A/B Testing Platform

disconnect_experimentation
Action · mcp:write

Disconnect a project's A/B-testing platform: deletes your stored (encrypted) credential for that vendor immediately. Only the account that connected it can disconnect it. query_experiments stops working for that vendor until connect_experimentation is run again.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "vendor": {
      "enum": [
        "intelligems",
        "vwo",
        "optimizely",
        "growthbook"
      ],
      "type": "string"
    }
  },
  "required": [
    "project",
    "vendor"
  ],
  "type": "object"
}

Connect Mixpanel

connect_mixpanel
Action · mcp:write

Connect a project's Mixpanel. Two paths: auth='oauth' returns an authorization URL for the user to open and approve (their org admin must have Mixpanel's MCP enabled); auth='service_account' (default) takes a customer-minted service-account username + secret — call once with no credentials to get instructions. Credentials are validated against the live Mixpanel API before being stored encrypted, and the connection binds to your token's email. Read-only usage only. Pass mixpanel_project_id when the credential sees multiple projects; region is us (default), eu, or in.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "auth": {
      "enum": [
        "service_account",
        "oauth"
      ],
      "type": "string"
    },
    "mixpanel_project_id": {
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "region": {
      "enum": [
        "us",
        "eu",
        "in"
      ],
      "type": "string"
    },
    "secret": {
      "description": "Service-account secret.",
      "type": "string"
    },
    "username": {
      "description": "Service-account username.",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Query Mixpanel
query_mixpanel
Read · mcp:read
READ-ONLY reads from the project's connected Mixpanel (connect_mixpanel first). Service-account connections use the Query API: action=events (event names), top_events, segmentation (params: event, from_date, to_date, unit), funnels (saved funnel list), funnel (params: funnel_id, from_date, to_date), retention (params: from_date, to_date, born_event). OAuth connections use Mixpanel's MCP: action=list_tools enumerates their 40+ analysis tools, action=call runs one (params: tool + arguments); write-shaped tools are refused.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "See description; default events / list_tools.",
      "type": "string"
    },
    "params": {
      "additionalProperties": true,
      "description": "Action parameters (from_date, event, funnel_id, tool, arguments...).",
      "type": "object"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Disconnect Mixpanel

disconnect_mixpanel
Action · mcp:write

Disconnect a project's Mixpanel: deletes your stored (encrypted) credential immediately. Only the account that connected it can disconnect it.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Connect Amplitude
connect_amplitude
Action · mcp:write
Connect a project's Amplitude. Preferred: auth='oauth' (default) returns an authorization URL for the user to open and approve in a browser; the grant is Amplitude's own mcp:read scope, so no key ever leaves their account. Fallback: auth='api_key' takes the project's API key + secret key from Settings → Projects (Amplitude has no read-only key variant; we only issue reads and store the pair encrypted). Call with auth='api_key' and no credentials to get instructions. Credentials are validated against the live Amplitude API before being stored, and the connection binds to your token's email. The key pair is project-scoped, so there is nothing to pick; pass amplitude_project_id only as a label. region is us (default) or eu.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "amplitude_project_id": {
      "type": "string"
    },
    "api_key": {
      "description": "Amplitude project API key.",
      "type": "string"
    },
    "auth": {
      "enum": [
        "api_key",
        "oauth"
      ],
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "region": {
      "enum": [
        "us",
        "eu"
      ],
      "type": "string"
    },
    "secret_key": {
      "description": "Amplitude project secret key.",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```
Query Amplitude
query_amplitude
Read · mcp:read
READ-ONLY reads from the project's connected Amplitude (connect_amplitude first). API-key connections use the Dashboard API: action=events (event catalog), segmentation (params: event, start, end, interval, metric), funnel (params: events list in order, start, end), retention (params: start_event, return_event, start, end), active_users (params: metric active|new, interval 1|7|30, start, end), sessions (params: start, end). Dates YYYY-MM-DD or YYYYMMDD, default last 30 days. OAuth connections use Amplitude's MCP: action=list_tools enumerates their tools, action=call runs one (params: tool + arguments); write-shaped tools are refused.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "See description; default events / list_tools.",
      "type": "string"
    },
    "params": {
      "additionalProperties": true,
      "description": "Action parameters (event, events, start, end, tool, arguments...).",
      "type": "object"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Disconnect Amplitude

disconnect_amplitude
Action · mcp:write

Disconnect a project's Amplitude: deletes your stored (encrypted) credential immediately. Only the account that connected it can disconnect it.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Connect Adobe Analytics

connect_adobe_analytics
Action · mcp:write

Connect a project's Adobe Analytics with an OAuth Server-to-Server credential the user's Adobe admin creates in their own Developer Console (client_id + client_secret + the scopes line). Call once with no credentials to get the instructions. The credential is validated live (IMS token, discovery/me, report suites) before being stored encrypted, and the connection binds to your token's email. Pass adobe_company_id when the credential sees several companies and rsid when the company has several report suites. Read-only usage only; Adobe has no browser-approval path Mimetic can host.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "adobe_company_id": {
      "description": "globalCompanyId, only if several.",
      "type": "string"
    },
    "client_id": {
      "type": "string"
    },
    "client_secret": {
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "rsid": {
      "description": "Report suite id, only if several.",
      "type": "string"
    },
    "scopes": {
      "description": "The Scopes line from the credential (optional; Adobe default used otherwise).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Query Adobe Analytics
query_adobe_analytics
Read · mcp:read
READ-ONLY reads from the project's connected Adobe Analytics (connect first). action=kpi (daily visits, visitors, page views, orders, revenue, bounces; params: start, end, metrics), top (params: dimension e.g. page, referringdomain, evar1; metrics; limit; start; end), metrics / dimensions (the report suite's catalog), segments, suites, report (params.body = a full Analytics 2.0 /reports request). Dates YYYY-MM-DD, default last 30 days, end exclusive. Adobe throttles at 12 requests per 6 seconds.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "action": {
      "description": "See description; default kpi.",
      "type": "string"
    },
    "params": {
      "additionalProperties": true,
      "description": "Action parameters (start, end, metrics, dimension, limit, rsid, body).",
      "type": "object"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```
Disconnect Adobe Analytics
disconnect_adobe_analytics
Action · mcp:write
Disconnect a project's Adobe Analytics: deletes your stored (encrypted) credential immediately. Only the account that connected it can disconnect it.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```
Connect PostHog
connect_posthog
Action · mcp:write
Connect a project's PostHog over OAuth. Returns an authorization URL for the user to open in a browser; approving grants READ-ONLY scopes (query:read, project:read). The connection binds to your token's email and picks the PostHog project matching the domain. After approving, query with query_posthog. Reconnecting re-mints tokens and rebinds the project.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Query PostHog

query_posthog
Read · mcp:read

Run a READ-ONLY HogQL query against the project's connected PostHog — events, unique users, funnels-by-hand, properties, session counts. Accepts a full HogQL SELECT; defaults to daily events + unique users over the last 14 days. Requires connect_posthog first.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "hogql": {
      "description": "HogQL SELECT, e.g. SELECT event, count() FROM events WHERE timestamp \u003e now() - INTERVAL 7 DAY GROUP BY event ORDER BY count() DESC.",
      "type": "string"
    },
    "limit": {
      "maximum": 1000,
      "minimum": 1,
      "type": "integer"
    },
    "posthog_project_id": {
      "description": "Override the bound PostHog project id.",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Connect Klaviyo
connect_klaviyo
Action · mcp:write
Connect a project's Klaviyo over OAuth. Returns an authorization URL for the user to open in a browser; approving grants READ-ONLY scopes (campaigns/flows/lists/segments/metrics). The connection binds to your token's email. If the site has NO email platform yet, use the Klaviyo signup link from get_connection_recommendations first, then connect.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Query Klaviyo

query_klaviyo
Read · mcp:read

READ-ONLY queries against the project's connected Klaviyo. Listings: campaigns, flows, lists, segments, metrics. Reports and checks: campaign_performance / flow_performance (opens, clicks, revenue attributed to Placed Order, per campaign/flow over a timeframe — sorted by revenue), abandoned_cart_status (is a cart/checkout-recovery flow live? the single highest-value check), dead_links (fetches recent campaign email HTML and HTTP-checks every real link; template-variable and unsubscribe links skipped). Requires connect_klaviyo first.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "filter": {
      "description": "Optional Klaviyo filter expression for listings, e.g. equals(status,\u0027live\u0027).",
      "type": "string"
    },
    "page_size": {
      "maximum": 100,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "resource": {
      "description": "What to fetch. Default campaigns.",
      "enum": [
        "campaigns",
        "flows",
        "lists",
        "segments",
        "metrics",
        "campaign_performance",
        "flow_performance",
        "abandoned_cart_status",
        "dead_links"
      ],
      "type": "string"
    },
    "timeframe": {
      "description": "For the performance reports. Default last_30_days.",
      "enum": [
        "last_7_days",
        "last_30_days",
        "last_90_days",
        "last_365_days",
        "this_month",
        "last_month"
      ],
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Connect Webflow
connect_webflow
Action · mcp:write
Connect a project's Webflow site over OAuth. Returns an authorization URL for the user to open in a browser; approving grants READ-ONLY access (sites:read, which also covers Webflow Analyze traffic reports). The connection binds to your token's email and covers every site the Webflow account authorizes. After approving, query with query_webflow.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Query Webflow

query_webflow
Read · mcp:read

READ-ONLY queries against the project's connected Webflow. Resources: sites (every site on the connected account: id, name, published domains, last published), traffic (Webflow Analyze daily sessions for the project's site over the last N days; needs the workspace's Analyze add-on, reported as not_enabled otherwise). Requires connect_webflow first.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "days": {
      "description": "traffic only: lookback window, default 30.",
      "maximum": 365,
      "minimum": 1,
      "type": "integer"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "resource": {
      "description": "Defaults to sites.",
      "enum": [
        "sites",
        "traffic"
      ],
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Connect Shopify
connect_shopify
Action · mcp:write
Connect a project's Shopify store for read-only queries. Checks whether the Mimetic Shopify app is installed and whether your token's email may read the store. Not installed -> returns install_url for the merchant (pass `shop` = the *.myshopify.com domain for a link that lands on the grant screen). Installed but your account is not yet approved -> returns approve_url: have the merchant open it inside Shopify admin and click Approve (15-minute, single-use), then call connect_shopify again. Alternatively the merchant can generate a code on the app's Claude Code page and you pass it as `code`. Then query with query_shopify.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "code": {
      "description": "Optional link code (XXXX-XXXX) the merchant generated inside the Mimetic Shopify app.",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "shop": {
      "description": "Optional *.myshopify.com domain, for a direct install link.",
      "type": "string"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}
```

Query Shopify

query_shopify
Read · mcp:read

READ-ONLY Shopify Admin GraphQL over the project's store, through the installed Mimetic app (orders, products, inventory, customers-free reports, checkouts, returns, themes: whatever the app's read scopes cover). Accepts a full GraphQL `query` document (mutations are rejected before they reach Shopify) plus `variables`; defaults to the shop identity + the last 30 days of orders. Requires connect_shopify first, and the token's email must be the store owner.

Input schema

{
  "additionalProperties": false,
  "properties": {
    "graphql": {
      "description": "Admin API GraphQL query document, e.g. { products(first: 10, sortKey: UPDATED_AT, reverse: true) { edges { node { id title totalInventory } } } }",
      "type": "string"
    },
    "project": {
      "description": "Project/site key (from list_projects).",
      "type": "string"
    },
    "variables": {
      "description": "GraphQL variables for the query.",
      "type": "object"
    }
  },
  "required": [
    "project"
  ],
  "type": "object"
}

Billing Status
billing_status
Read · mcp:read
Return the Mimetic billing status for a project: plan, whether it is on a paid/active plan, and checkout/portal URLs.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}
```
Get Checkout Link
get_checkout_link
Action · mcp:write
Return a Stripe Checkout URL to upgrade a project to a paid Mimetic plan. Use when the user wants to pay or upgrade: give them the URL to open in a browser. They enter payment details on Stripe and the plan activates automatically once payment completes.
Input schema
```
{
  "additionalProperties": false,
  "properties": {
    "plan": {
      "description": "Plan to check out (default starter).",
      "type": "string"
    },
    "project": {
      "type": "string"
    }
  },
  "type": "object"
}
```

Your site’s growth data,in your AI assistant.

Connect to Mimetic

Remote MCP

CLI and local MCP clients

What to ask

Find revenue opportunities

Investigate landing-page performance

Use experiment results

Review visitor friction

Set up missing measurement

Example tool request

Access and permissions

Live tool catalog

Your site’s growth data,
in your AI assistant.