21RISK is in Claude's connector directory, and this is its MCP server address for Claude Code and any other MCP client. Connect, sign in, and your assistant answers from live 21RISK data, within the access you grant.
https:/No API key and no plug-in: the URL and your 21RISK login are all it takes.
Every guide ends at the 21RISK sign-in, where you choose which organizations the assistant may reach and whether it may change anything.
21RISK is listed in Claude's connector directory. Connect it once, and it is there in every chat.
Open the 21RISK listing in Claude's connector directory. Or find it yourself: Customize → Connectors → Discover, and search for 21RISK.
Open 21RISK in ClaudeSelect Connect to Claude. The 21RISK sign-in opens: use your normal login, single sign-on included.
Tick the organizations Claude may reach, pick Read-only access or Full access for each, and click Allow.
In a chat, select + → Connectors and check that 21RISK is on. Then ask about your sites, values or open actions.
One command in your terminal, then a sign-in from inside Claude Code.
--scope user makes 21RISK available in every project. Leave it
out to add it to the current project only.
claude mcp add --transport http --scope user 21risk https:/Start claude, run /mcp, select 21risk and choose to authenticate. The 21RISK sign-in opens in
your browser; pick the organizations and access there. From the shell, claude mcp login 21risk does the same.
Ask in plain language. Claude Code looks up the right 21RISK function and runs it with the access you granted.
{
"mcpServers": {
"21risk": {
"type": "http",
"url": "https:/21RISK speaks standard remote MCP with OAuth, so any client that supports remote servers can connect: VS Code, for example.
Add an MCP server of type HTTP (some clients say Streamable HTTP) with the server URL. Leave headers and API keys empty.
https:/On first connect the client finds the 21RISK sign-in on its own and opens your browser. Sign in, choose organizations and access, and click Allow.
The client lists the 21RISK tools; ask your question.
{
"mcpServers": {
"21risk": {
"type": "http",
"url": "https:/Nothing is shared until you click Allow on the 21RISK authorization screen.
The assistant works as your 21RISK user. It sees only what you can see, and what it does is logged under your name.
Read-only can look anything up and change nothing. Start there: you can raise it later without reconnecting.
Every connected assistant is listed under Settings → Profile. Revoke one and it loses access on its next request.
No. Every connection is a person signing in with OAuth, the standard behind "Sign in with…" buttons. The assistant never sees your 21RISK password, and its tokens are short-lived and refresh while the connection is active.
The MCP server works on every plan, with fair-use rate limits per plan.
Yes. The authorization screen lists every organization you belong to. Tick the ones to include and set read-only or full access for each.
No, it keeps working. To use the directory listing instead, remove the custom connector under Customize → Connectors, connect 21RISK from Discover, and sign in to 21RISK again.
In 21RISK, open Settings → Profile and find Authorized Applications. Revoke the assistant there and it loses access on its very next request.
The documentation covers the access model in depth. New to 21RISK? Sign up first, then connect.