MCP servers for goose
goose is an open-source, on-machine AI agent from Block that runs on whichever model you point it at, and every goose extension is an MCP server.
Connect BusinessMCP to goose
One entry for every tool you have connected, rather than one config block per server with its own credential. Add this to ~/.config/goose/config.yaml:
- 1Open ~/.config/goose/config.yaml (or use `goose configure` and pick Add Extension → Remote Extension).
- 2Add the block above with a key from your BusinessMCP dashboard.
- 3Restart goose. The tools appear alongside its built-in extensions.
- 4Ask it something about your own numbers — it calls the tool instead of guessing.
Servers worth connecting to goose
Official and verified entries from the directory. Each links to a detail page with its real tool list, example prompts and setup steps.
Stripe
Stripe, Inc.
Payments, subscriptions, invoices, and revenue analytics as agent-callable tools.
GitHub
GitHub
Repos, PRs, issues, and Actions — controllable from any agent thread.
Salesforce
Salesforce
Leads, opportunities, and pipeline state for outreach and sales agents.
Google Search Console
Query data, index coverage, and sitemap tools for SEO agent threads.
Meta Ads
Meta
Campaign creation, budgets, and creative deployment via the Meta Ads API.
Google Analytics
User counts, funnels, and traffic sources for your dashboard widgets.
Slack
Slack
Post updates, read channels, and bridge agent threads into your workspace.
free directory submitter
BusinessMCP
Automatically drafts and registers your startup to launch directories.
seo onpage auditor
BusinessMCP
Scrapes any URL and returns structured, prioritized SEO recommendations.
Frequently asked questions
Is a goose extension the same thing as an MCP server?
Yes. goose calls them extensions, and they speak MCP — which is why a hosted endpoint works without writing any goose-specific code. It moved to the Linux Foundation’s Agentic AI Foundation in April 2026.
Do the charts render in goose?
Yes. goose is one of the hosts that supports MCP Apps, so a result from an analytics tool renders as an actual chart inline rather than as JSON the model paraphrases. Claude Desktop and VS Code Copilot do the same.
Which model should I run it with?
Whichever you like — that is goose’s point, and ours: the endpoint is a protocol, so the tools and the access policy are identical whatever is on the other end.
One endpoint, every tool, in goose
Connect your stack once and it is reachable from goose — and from every other AI you use — without a config block per server.
Get started freeFree plan available · No credit card