SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI marketingmureo

mureo

Local-first framework for operating Google Ads, Meta Ads and Search Console, grounded in a local STRATEGY.md file.

Open website

About this tool

mureo is an open-source, local-first AI advertising operations control plane, positioned as an "AI advertising operations team." It lets AI agents (Claude Code, Cursor, Codex, Gemini, etc.) operate Google Ads, Meta Ads, Amazon Ads, TikTok Ads, Search Console, and GA4 through mureo, while ensuring every action is based on business strategy, tied to real results, and recorded in a replayable audit log. Credentials always remain on the user's local machine.

Main capabilities

  • Strategy-driven: Every decision reads the local STRATEGY.md (user personas, unique selling points, brand tone, goals), so agents optimize around business goals rather than only optimizing metrics.
  • Safety gating: Includes a rollback allowlist, GAQL input validation, BYOD read-only by default, credential guards, per-platform rate limiting, and other mechanisms.
  • Cross-platform: Orchestrates Google Ads, Meta Ads, Amazon Ads, TikTok Ads, Search Console, and GA4 in a single workflow.
  • Auditable: An append-only action log supports rollback and records every tool call and its result (mureo journal).
  • Local-first: Credentials never leave the local machine, and mureo itself has no telemetry.
  • Learnable: /learn builds an account-specific knowledge base over time.

How to use

Quick start (demo)

Requires Python 3.10+ and Claude Code (Cursor, Codex CLI, and Gemini CLI are also supported). The demo scenario uses synthetic data and requires no ad account credentials, OAuth, or sign-up. Run mureo configure to open the local browser UI (bound to 127.0.0.1), select the Claude app, complete one-click basic setup, then choose a demo scenario in the Demo / BYOD section. Terminal equivalent command:mureo setup claude-code --skip-auth && mureo demo init --scenario seasonality-trap. After that, open the generated demo directory in Claude Code and ask questions; the agent will read the demo STRATEGY.md, pull campaign data, and identify seasonal traps that numeric tools alone would miss.

Path A: Bring your own data (BYOD)

Export real account data as XLSX, import it into mureo, and get strategy-based multi-platform diagnostics without OAuth or API access approval. Import from the mureo configure dashboard or the terminal. The first run of /onboard generates STRATEGY.md and STATE.json through an interview. BYOD is read-only by construction: all mutation tools return {"status": "skipped_in_byod_readonly"}, the agent only analyzes and advises, and does not write to the account.

Path B: Direct connection (OAuth)

Connect mureo directly to the Google Ads / Meta Ads APIs to actually execute changes (such as /rescue, /budget-rebalance, /creative-refresh, or apply rollbacks via rollback_apply), with support for GA4 / Search Console. Complete interactive OAuth in Connect platforms in the mureo configure UI. Requires a Google Cloud project (including Google Ads API access and an OAuth client) and/or a Meta App ID + Secret (development mode is sufficient). After connecting, run /onboard once to generate STRATEGY.md and STATE.json.

Use cases

  • Daily health checks for ad accounts, search term cleanup, competitor scans, budget rebalancing, creative refresh, and more.
  • Cross-platform performance correlation analysis to identify issues such as brand cannibalization and mismatched conversion definitions.
  • AI-driven ad operations that require audit trails and rollback capabilities.
  • Teams or agencies can use it through the commercial edition (cloud-hosted service or on-premises Agency edition); see mureo.jp for details.

Limitations

  • The demo and BYOD modes are read-only and cannot actually execute changes; executing changes requires a direct OAuth connection.
  • OAuth setup requires experience configuring Google Cloud Console or Meta for Developers.
  • Amazon Ads is bridged through the official MCP, and credentials do not enter the host MCP configuration; the official TikTok Ads MCP is already supported.
  • Specific fees, pricing, and the complete list of platform coverage are not provided in the materials.

Security design

  • Credential guard: installs a PreToolUse hook to prevent the AI agent from reading secrets files such as ~/.mureo/credentials.json and .env.
  • GAQL input validation: all IDs, dates, and string literals that enter Google Ads queries undergo allowlist validation.
  • Anomaly detection: compares current metrics against a median-based baseline and issues alerts for zero spend, CPA spikes, CTR drops,and more.
  • Rollback allowlist: only operations on the allowlist can have rollbacks planned, and destructive verbs and unexpected parameter keys are rejected.
  • Immutable data model: all state objects are frozen=True data classes.
  • Local credentials: tokens are sent only to the official ad platform APIs.

MCP tools

mureo exposes 229 MCP tools via stdio: Google Ads (92), Meta Ads (90), Search Console (10), plus rollback, bulk, change import, anomaly detection, delivery collapse detection and diagnosis, exclusion delivery impact preview, tracking parameter consistency, strategy/state context, analytics registry, learning, learning period reset precheck, and Creative Studio. After configuring Amazon Ads, bridged Amazon tools are added from the local manifest. Any MCP-compatible client can connect.

Related tools

Browse category