Connect a Model Context Protocol server, discover what it actually offers and choose which employee can use each tool. Every external action follows the company’s permission and approval path.
These are available connection options. No account is connected on your behalf, and a preset does not guarantee access to every provider tool.
Repository, issue, and pull-request work through GitHub's official server.
Supply a narrowly scoped GitHub token in an MCP_ environment variable. Repository access and organisation policy still apply. Enable only the tools each employee needs.
Provider setup documentation ↗Browser interaction through a separately operated local Playwright MCP server.
Start the official Playwright MCP server separately on loopback port 8931 using an isolated browser profile. Browser actions can submit forms or change accounts; grant them deliberately. This application does not launch a server or inherit your browser login.
Provider setup documentation ↗Design context and available canvas tools from Figma's remote server.
Figma currently accepts only supported clients in its MCP Catalog. Nucovo needs Figma client approval before this preset can connect. Plan and file permissions also affect available tools.
Provider setup documentation ↗Workspace knowledge and page operations through Notion's hosted MCP server.
Connect a Notion account and select the workspace access it may use. Only discovered and explicitly enabled tools become available to employees.
Provider setup documentation ↗Jira and Confluence work through Atlassian's Rovo MCP server.
Authorise an Atlassian account with access to the intended cloud sites. Organisation policies and product permissions still restrict the available actions.
Provider setup documentation ↗Canva's design connector for design creation, editing, search, and export.
This is Canva's design MCP, not Canva Dev documentation tools. Custom clients must obtain redirect-URI allowlisting and each Canva user must authorise access. Some tools depend on plan entitlements.
Provider setup documentation ↗Enter a Streamable HTTP endpoint with its authentication and approved OAuth destinations. Use HTTPS for remote connections; local HTTP is allowed only when loopback access is explicitly enabled. Tool discovery determines what is available.
This connection does not launch a local process or inherit tools from your personal Claude Code, Codex or Gemini CLI configuration. Other transports and arbitrary private-network endpoints are not included.
Create a design or marketing project with the intended dimensions, audience, brand material and expected outputs. Connect Canva, discover the available design tools and grant the Designer only what the project needs.
Read access can help the employee locate source material. Creation, edits and exports use the discovered tool schemas and require the configured approvals. The final handoff must distinguish a written design brief from a design or export actually returned by Canva.
This uses Canva’s design MCP, not a documentation-only developer server. A custom client needs Canva’s redirect-URI allowlisting and the account holder’s consent; some capabilities depend on their plan. No native asset creation is promised before that setup succeeds.
The console lists tool descriptions and input schemas returned by the server. Grants bind to the configuration version and tool definition; material changes invalidate old access and require another review.
Grant a specific discovered tool to a named employee. Automatic access is limited to tools the operator has reviewed as read-only. A server’s read-only hint alone is not that review.
Other calls wait for approval tied to the exact request. Arguments are validated against the supported schema before dispatch. Approval permits a call; it does not prove that the remote service will behave safely or that the result is correct.
Call records connect the employee, task, tool and outcome. A write with an uncertain result is not automatically replayed. The operator must check what happened at the service before trying again.
OAuth credentials are stored encrypted by the trusted application process, using an installation key. Tokens stay outside model prompts and the code runner. Tool arguments and results may contain project information; grant access only to suitable destinations. External results remain untrusted input. Read the MCP security boundaries →
Start in Company → MCP Connectors. Configure an endpoint, authorize the account, inspect its tools and assign the access your employees need. Provider account policies, schema support and available tools determine what can run.
Nucovo