MCP
Boldo exposes a Model Context Protocol server so MCP-compatible assistants can query your Boldo data and documentation.
Use MCP when you want an external assistant to inspect your architecture context while keeping access scoped to a selected organization and to the user's own Boldo rights.
MCP access is read-only. Connected assistants can query data, but they cannot modify your Boldo knowledge base.
Access
MCP access depends on your plan, organization AI availability, and the user's AI access.
Go to Organization → Developer. When MCP is available, the page shows the MCP connection card and the MCP documentation link.
Connect a client
- Open Organization → Developer
- Open the MCP documentation from the MCP card
- Copy the MCP server URL
- Add it to your MCP-compatible client configuration
- Start the OAuth connection from your client
- Select the Boldo organization to authorize
- Approve the connection only if you initiated it yourself
The organization is selected during OAuth consent. The same MCP server URL can be used before selecting the organization.
Only organizations that the current session can access are available for authorization. If an organization requires SSO, start the OAuth flow from a session authenticated with the matching SSO provider.
What MCP can query
The MCP server exposes tools for the main read use cases:
| Area | Typical use |
|---|---|
| Assets | Search, inspect, and resolve assets |
| Metamodel | Read asset types, properties, and relationship types |
| Diagrams | Inspect diagram content |
| Charts | Inspect chart configuration and context |
| Views | Inspect saved views |
| Nested maps | Inspect nested map structure |
| Process flows | Inspect process flow context |
| Documentation | Search Boldo documentation |
The interactive MCP documentation lists the exact tools and parameters available in the current version.
Process flow tools require a plan that includes process flows. Without that plan, process flow tools are not exposed.
The exact tools also depend on the organization AI access level. With Metamodel, MCP can expose structure and documentation tools. With Data, it can also expose read tools for data the approving user is allowed to see.
Permissions
MCP acts on behalf of the user who approved the OAuth connection.
That means:
- data access follows the user's Boldo rights
- the selected organization scopes the connection
- the user's AI access and the organization's AI/MCP availability are checked
- SSO requirements, enabled sign-in methods, 2FA enforcement, and IP restrictions still apply
- revoking a connection stops that client from using the authorization
- an administrator can revoke organization-level connected apps
MCP authorizations are checked when the client calls Boldo. If the user loses access, the organization disables MCP or AI, or an administrator revokes the connected app, the client stops working automatically.
Boldo also rate-limits MCP requests. A client that exceeds the limit receives a temporary error and should retry after a short wait.
Manage connected apps
Users can review and revoke their own connected MCP apps from My account → Security.
Administrators and owners can review organization-level connected apps from Organization → Developer when MCP is enabled. Editors with Developer access do not see this section.
Revoking access affects the selected organization connection. Other members and other organizations are not automatically affected.
API keys vs MCP
API keys and MCP serve different integration needs.
| Need | Use |
|---|---|
| Build scripts, automations, or system integrations | API |
| Let an assistant inspect Boldo through a conversational client | MCP |
| Write or update Boldo data programmatically | API |
| Give read-only contextual access to an AI agent | MCP |
Use an API key when the integration must write data or run controlled backend jobs. Use MCP when an assistant needs read-only context for analysis and conversation.