# Local MCP for agent workflows while the app is active. > See premium local foreground MCP for text, URL, file, and repository scans, redacted reports, and runtime guards. - Canonical page: [https://veridicuscan.app/mcp-automation](https://veridicuscan.app/mcp-automation) - Language: English Premium MCP mode Veridicus Scan includes a premium, local, foreground, session-based MCP surface. Use it when an agent should run `scan_text`, `scan_url`, `scan_file`, or `scan_repo`, retrieve a report, or guard a risky plan or action while the app runtime is active. MCP guardrails - Premium entitlement required for session opening, scan, report, and runtime-guard methods - `health`, `list_methods`, and `close_session` stay callable without premium - Only one active session is allowed at a time - Built for local foreground use, not an always-on daemon Core flow ## The MCP path is explicit and session-based. The bridge does not hide state. Open a session, run the local tools you need, then close the session. That keeps the automation surface close to the app’s trust model instead of turning it into an unbounded background service. 01 ### Open a session Start with `open_session` and carry the returned session ID through every stateful MCP method. 02 ### Run scans or guardrail methods Choose `scan_text`, `scan_url`, `scan_file`, or `scan_repo`, then use report or runtime-guard methods needed for the current task. 03 ### Close the session cleanly Call `close_session` when the workflow finishes. The current local policy allows one active session at a time. Method families ## The MCP feature goes beyond scan-and-return. The bridge exposes source-specific scan methods, report surfaces, and runtime guards. Method names stay explicit so an agent can choose the narrowest input path and preserve the returned policy scope. ### Scan sources - `scan_text` for prompts, snippets, or extracted text - `scan_url` for a user-chosen public HTTPS destination - `scan_file` for one local artifact - `scan_repo` for bounded repository intake ### Reports and context - `get_report` for the current reviewed result - `export_report` for JSON or PDF handoff - Evidence snippets are redacted by default on the applicable sanitized report path - Coverage notes travel with bounded or partial results ### Runtime guardrails - `scope_tools` establishes the authoritative tool scope - `guard_plan` evaluates a proposed plan - `gate_action` evaluates a specific tool action What it is and is not ## MCP adds automation without turning the app into a hosted gateway. MCP lets Veridicus Scan participate in agent execution and memory flows while keeping the same local runtime boundary: foreground, session-based, and designed to stay inside the app instead of becoming an always-on remote service. Local ### Runs inside the local foreground model The bridge is designed for on-device use while the app or local runtime is active, not as a permanent hosted gateway. Bounded ### Repository scanning is an intake guard, not a complete verdict `scan_repo` uses bounded traversal and repo-aware signals before an agent reads or installs from a repository. It does not claim to replace a full code or dependency security review. Premium ### MCP is treated as a premium capability The core MCP methods are premium-protected because they turn the app into an active agent-runtime component. Entitlement and localized pricing are handled through StoreKit rather than a hardcoded site price. Next step ## Bring MCP into agent workflows, trust decisions, and exported reports. Move from the automation surface into use cases, MCP security guidance, best-practices guidance, trust boundaries, exported evidence, or OpenClaw intake safety.