Skip to main content
Community Plugins let the Adspirer community extend the product with extra analysis and insight tools that show up right alongside the built-in ones. They’re open source, they’re read-only to your ad accounts, and they run in a locked-down sandbox — so you get more capability without new risk. Your AI client reaches them through a single community_plugins router tool: it calls action: "list_tools" to see what’s available, then action: "execute" to run one.

​
What they are

  • Open source. Every plugin lives in the public Adspirer/adspirer-mcp-community repository (Apache-2.0). Anyone can read the code, contribute a plugin, or fork one.
  • Analysis, not automation. Plugins cluster, score, rank, format, and recommend on top of your ad data. A first example is a keyword clusterer (cluster_keywords) that groups your search terms into themed clusters.
  • Reviewed and pinned. A plugin only becomes available after an Adspirer maintainer reviews and merges it, and the server loads it from a pinned commit — so the code you run is the exact code that was reviewed.

​
Why they’re safe

Community plugins are read-only to your ad accounts (v1). They can analyze and recommend, but they cannot create, edit, pause, or delete anything in your campaigns.
  • Credential-free & sandboxed. Plugins run in a separate, egress-locked sub-server. They never see your OAuth tokens or credentials.
  • Read-only broker. They reach your data only through a narrow read-only interface (list campaigns, get performance, analyze search terms, check connection status) — never a write path.
  • Import-gated. A load-time check rejects any plugin that tries to import something outside the allowed set.
  • Separately metered. Community-plugin calls draw on their own allowance, distinct from your core tool-call quota, so experimenting with them doesn’t eat into your campaign-management budget.

​
Using a plugin

Discover and run a community analysis plugin.

Open in Cursor

​
Contributing a plugin

Community plugins are a great way for agencies and power users to package the analysis they run over and over. Add a Python MCP analysis tool to the public repository, following the contribution guide; once a maintainer reviews and merges it and the commit is pinned, it appears for everyone.

​
FAQ

No. Community plugins are read-only to your ad accounts in v1 — they analyze and recommend only. They run in a sandbox with no credentials and no write path, so there is no way for a plugin to alter a campaign or spend budget.
Community plugins are open-source tools (extra analysis capabilities) shared through a public repo. Community playbooks are shareable agent strategies — the goal, rules, and guardrails of an Ad Ops agent, with no account data attached. Plugins add capabilities; playbooks add strategy.
No. Approved community plugins are available through the community_plugins tool on the hosted server — your AI client discovers them automatically. There’s nothing to install locally.

Autonomous Ad Ops Agents

Background agents that propose changes to an approval queue.

Tool Catalog

The full 400+ built-in tool reference.
Last modified on July 8, 2026