Meta, Instagram, Facebook Ads MCP

armavita-meta-ads-mcp is a local Model Context Protocol server for Meta Ads. It is built for local MCP clients (Claude Code, Cursor, Codex) and supports:

• Meta access token auth (META_ACCESS_TOKEN)

• Local OAuth flow (META_APP_ID + META_APP_SECRET)

• stdio MCP transport only

• Python 3.11+

• mcp[cli]==1.27.2

• Meta Marketing API v25.0 by default (META_GRAPH_API_VERSION override supported)

Current contract version: 1.2.0 (78 tools).

The default OAuth scope (META_AUTH_SCOPE) requests ads_management, ads_read, business_management, public_profile, pages_show_list, pages_read_engagement, instagram_basic, and threads_business_basic. The catalog tools additionally require catalog_management — add it to META_AUTH_SCOPE if you use them.

Install

From PyPI (once published):

pip install armavita-meta-ads-mcp

From source (recommended during development):

uv sync

Related MCP server: Meta Ads MCP

Run

armavita-meta-ads-mcp

Module entrypoint:

python -m armavita_meta_ads_mcp

Login flow:

armavita-meta-ads-mcp --login

Quick MCP Client Config

Minimal MCP server registration (JSON format used by many clients):

{
  "mcpServers": {
    "meta-ads-armavita": {
      "command": "armavita-meta-ads-mcp",
      "env": {
        "META_ACCESS_TOKEN": "EA...",
        "META_GRAPH_API_VERSION": "v25.0"
      }
    }
  }
}

OAuth mode (no direct token in config):

{
  "mcpServers": {
    "meta-ads-armavita": {
      "command": "armavita-meta-ads-mcp",
      "env": {
        "META_APP_ID": "YOUR_APP_ID",
        "META_APP_SECRET": "YOUR_APP_SECRET"
      }
    }
  }
}

Then run once to complete login:

armavita-meta-ads-mcp --login

Advanced Environment Variables

Beyond the core auth variables above, the server reads four optional knobs:

Tool Coverage (78 tools)

Accounts & controls

• list_ad_accounts, read_ad_account, get_account_controls, update_account_controls, list_account_pages, search_pages

Campaigns

• list_campaigns, read_campaign, create_campaign, update_campaign, create_campaign_budget_schedule

Ad sets

• list_ad_sets, read_ad_set, create_ad_set, update_ad_set

Ads, creatives & media

• list_ads, read_ad, create_ad, update_ad, list_ad_previews

• list_ad_creatives, read_ad_creative, create_ad_creative, update_ad_creative

• upload_ad_image_asset, upload_ad_video_asset, read_ad_image, export_ad_image_file

• list_ad_images, list_ad_videos

Insights & reporting

• list_insights, create_report

• Async insights: create_insights_job, read_insights_job, read_insights_job_results

• list_ad_custom_derived_metrics

Custom conversions

• list_custom_conversions, read_custom_conversion, create_custom_conversion, update_custom_conversion, delete_custom_conversion

Targeting research

• search_interests, suggest_interests, estimate_audience_size, search_behaviors, search_demographics, search_geo_locations

Audiences

• list_custom_audiences, read_custom_audience, create_custom_audience, update_custom_audience, delete_custom_audience

• create_lookalike_audience, manage_custom_audience_users

Catalogs & products

• list_product_catalogs, list_product_sets, list_products, upsert_product, batch_products

Duplication

• clone_campaign, clone_ad_set, clone_ad, clone_ad_creative

Ads Library & research helpers

• search_ads_archive

• search_web_content, read_web_content

Conversions API (CAPI)

• send_capi_events, list_business_datasets, read_dataset_quality

Partnership & branded content

• list_branded_content_ad_permissions, grant_branded_content_ad_permission

Reach & frequency

• list_reach_frequency_predictions, create_reach_frequency_prediction, read_reach_frequency_prediction

Recommendations

• list_recommendations, apply_recommendation

Threads

• create_threads_account, get_threads_account

Pagination

Cursor-based pagination is supported on list/read streams that expose page_cursor:

• Accounts: list_ad_accounts

• Campaigns: list_campaigns

• Ad sets: list_ad_sets

• Ads: list_ads, list_ad_creatives

• Media library: list_ad_images, list_ad_videos

• Insights: list_insights, read_insights_job_results

• Targeting: search_interests, suggest_interests, search_behaviors, search_demographics, search_geo_locations

• Audiences: list_custom_audiences

• Custom conversions: list_custom_conversions

• Catalogs: list_product_catalogs, list_product_sets, list_products

• Derived metrics: list_ad_custom_derived_metrics

• Partnership: list_branded_content_ad_permissions

• Reach & frequency: list_reach_frequency_predictions

• CAPI datasets: list_business_datasets

• Ads Library: search_ads_archive

Use page_cursor with the paging.cursors.after value from the previous response. Responses preserve Meta's native paging object.

Insights Query Notes

• list_insights accepts an optional fields parameter (defaults to a standard KPI set). Custom derived metric names from list_ad_custom_derived_metrics can be passed in fields.

• list_insights and create_report support either:

  • date_range as { "since": "YYYY-MM-DD", "until": "YYYY-MM-DD" }, or

  • date_range as a preset (for example last_30d, maximum).

• create_report.comparison_period uses the same format and validation as date_range.

• list_insights, create_report, and async insights accept action_attribution_windows. Deprecated windows (7d_view, 28d_view) return warnings and may yield empty data under v25.

• previous_30d is normalized to last_30d.

• For action metrics, use action_breakdowns (and optional summary_action_breakdowns) instead of mixing action keys into breakdowns.

Async insights workflow (including MMM CSV)

1. create_insights_job(object_id, breakdowns=["mmm"], export_format="csv", ...) → report_run_id

2. Poll read_insights_job(report_run_id) until async_status is Job Completed (also returns async_percent_completion and async_report_url when available)

3. Fetch rows via read_insights_job_results(report_run_id) or download async_report_url

v25 behavior notes

• API version: default v25.0 — the current latest Marketing/Graph API version (released Feb 18, 2026). Set META_GRAPH_API_VERSION to override. Watch the Meta changelog before adopting a future version as the default.

• Campaign budgets: create_campaign no longer silently applies daily_budget=1000. Provide daily_budget/lifetime_budget, set use_ad_set_level_budgets=True, or pass apply_default_budget=True to opt in to the MCP default.

• Special ad categories: when special_ad_categories includes HOUSING, EMPLOYMENT, or FINANCIAL_PRODUCTS_SERVICES, special_ad_category_country is required on create/update.

• Ad set placement opt-out: placement_soft_opt_out must be an object keyed by placement group, for example:

  {
    "facebook_positions": ["marketplace"],
    "instagram_positions": ["stream"]
  }

  Allowed keys: facebook_positions, instagram_positions, audience_network_positions, messenger_positions, threads_positions.

• Advantage+ duplication/create guard: deprecated Advantage+ Shopping/App campaign signatures are blocked on create_campaign (when smart_promotion_type is set) and on clone_campaign / clone_ad_set / clone_ad preflight. Use migrate_to_advantage_plus on update_campaign / clone_campaign where supported.

• Carousel & catalog creatives: create_ad_creative supports carousel_cards, product_set_id (see list_product_sets), and url_tags. create_ad supports conversion_domain.

• Video upload: upload_ad_video_asset accepts video_source_url or local video_file_path (multipart).

• Ads Library: search_ads_archive exposes v25 filters including ad_active_status, delivery date bounds, search_page_ids, search_type, languages, media_type, and publisher_platforms.

Security

• Access tokens are redacted from URL fields returned by the server (including nested paging.next URLs).

Docs

• Authentication Setup Guide

• Local Client Guide

• Security Policy

Scope

• This repository is an OSS local MCP server.

• Transport mode is local stdio only.

• Tool aliases are intentionally not exposed.

License

GNU Affero General Public License v3.0 (AGPLv3). See LICENSE.