Every MCP server, governed by default.
Reach approved tools from Claude, Cursor, ChatGPT, and agents — one endpoint, scoped credentials, policy and an audit trail on every request. Every connector carries a risk tier, and risky ones stay gated until an admin approves — then every consequential call routes through Fairy's judgment engine.
Atlassian
medium5 tools · 4 resources · 1 prompts
Linear
medium16 tools · 3 resources · 2 prompts
Slack
high12 tools · 2 resources · 1 prompts — touches personal data
Perplexity
medium8 tools · 2 resources · 2 prompts
Shortcut
medium6 tools · 2 resources · 6 prompts
GitHub
medium22 tools · 6 resources · 3 prompts
Stripe Payments
high14 tools · 3 resources · 1 prompts — touches money
Salesforce
high20 tools · 8 resources · 2 prompts — touches personal data
Postgres Admin
critical9 tools · 1 resources · 0 prompts — destructive/admin surface, touches production/infra
Notion
high11 tools · 5 resources · 2 prompts — touches production/infra
Risk tier is computed on import from every federated registry — a payments or prod-admin server is flagged and gated automatically, not left to a reviewer to notice. That's the line: every MCP server, governed — not just listed.