Model Context Protocol
Use Frontal through MCP clients while preserving the same workspace, permission, and action boundaries.
The Frontal MCP integration makes selected platform capabilities available in compatible AI clients. It is useful when a person needs to query context, inspect a run, start approved work, or export a result without leaving their usual interface.
Access follows the request
An MCP client does not bypass Frontal permissions. Authenticate the connection to the correct workspace, grant only the tools the client needs, and keep writes behind the same approvals and policy checks used elsewhere.
Start read-only
Begin with entity lookup, run inspection, and reports. Add write-capable tools only after you have reviewed the traces and decided how approval, idempotency, and error recovery work in the client.
Client-specific setup belongs on the corresponding MCP page. See the public MCP directory for supported clients and connection details.
更新于 2026年8月31日