Use Vincia in your AI client
Vincia exposes a public MCP server at https://mcp.vincia.io. Add it to your LLM client and the catalog, validation, and webhook-testing tools appear in your chat — no install, no CLI, no auth.
One-click add
Pick your client. The buttons below open your client's connector flow with the URL pre-filled. If your browser doesn't handle the deep link, use the copy-button below and paste manually.
What you get
The public surface exposes 19 tools from Vincia's contributor MCP and 8 docs resources covering the contributor platform end-to-end.
Catalog & lookup
vincia:list_capabilities— the five SDK capabilitiesvincia:get_interface— TypeScript types for a capabilityvincia:get_scope_catalog— all permission scopesvincia:get_resource_caps— per-tier resource ceilingsvincia:get_contract_version— active platform contract versionvincia:find_examples— published example contributionsvincia:get_plugin_source— source of a published pluginvincia:search_marketplace— query the marketplace
Validation
vincia:lint_manifest— validate against the v1.0 schemavincia:check_scopes— detect required scopes from code
Webhook testing (chat-first)
vincia:create_webhook_inbox— get a durable URL Stripe / GitHub / OAuth callbacks can hitvincia:list_webhook_events— poll an inbox for arrivalsvincia:webhook_event_to_fixture— convert a captured event into a simulate fixture
Onboarding prompts
Six guided prompts (audit, scaffold, simulate, lint, search, publish) walk an LLM through a first contribution — useful as self-directed onboarding inside any client.
Docs resources
The MCP resources/list exposes 8 docs sections at vincia://docs/<slug>: preview map, sandbox previews, simulations, webhook inboxes, CLI reference, MCP tools, auth + rate limits, troubleshooting. Ask the LLM and it pulls the relevant section automatically.
What's NOT in the remote surface
Eight tools are stdio-only because they touch your local filesystem (workspace_dir, bundle_path) or invoke local subprocesses: simulate_run, scaffold_contribution, render_widget_preview, render_blueprint_preview, render_pack_preview, render_simulation, render_composition_preview, list_preview_drafts.
For those, install the Vincia CLI and add the stdio MCP server:
curl -fsSL https://get.vincia.io/install | bash vincia login # one-time, saves a session token vincia mcp install-claude-desktop # or --cursor / --windsurf / --vscode
With the stdio server installed, the same chat surface plus the local-filesystem tools light up.
Try these prompts
- Browse the catalog
Use the Vincia connector. List the SDK capabilities, then find me three payment-connector examples.
- Validate a manifest
Lint this connector manifest against Vincia's v1.0 contract: {paste your manifest.json here}.
- Look up the rules
What are the per-tier resource caps in Vincia? Show me T1 specifically.
- Test a webhook handler (chat-first)
Create a webhook inbox labelled 'stripe-test'. I'll paste the URL into Stripe's dashboard. When events arrive, convert one into a vincia simulate fixture so I can replay it.
Subscription tiers (heads-up, not a block)
- Claude.ai web — Connectors require a Pro or Team plan. Free users hit a paywall on the Connectors page.
- ChatGPT web — Custom MCP connectors require Plus or Pro AND Developer mode enabled under Settings → Apps → Advanced settings (ChatGPT marks third-party connectors as “ELEVATED RISK”; that's expected).
- Claude Desktop — Free tier supports remote MCP connectors. The smoothest free path.
- Cursor / Windsurf / VS Code Copilot Chat — All accept remote MCP URLs without a plan upgrade.
Server status
Liveness: https://mcp.vincia.io/health. Anonymous per-IP rate limits: 60 read-only tool calls per minute, 5 heavy calls per hour. Limits emit as tool-result envelopes (not HTTP 503) so your client renders them inline.