MCP Server
The agent asks Semgrep and Semgrep answers. The MCP server exposes Semgrep scanning as tools the agent can call directly.
Learn more about the MCP →Guardian runs your Semgrep policies the moment an agent writes a file, blocks what violates them, and returns the finding context in the session.
Claude Code › writes payments/api.py
+ query = f"SELECT * FROM users WHERE id = {user_id}"
Guardian hook › runs your Semgrep policy
acme-prod · SQL injection BLOCKED
User input reaches a database query without parameterization.
Claude Code › updates the query using the finding context
- f"… WHERE id = {user_id}"
+ cur.execute("… WHERE id = %s", (user_id,))
Guardian › policy check passed on the updated file
Guardian combines deterministic Semgrep checks with the context of the active Claude Code session. Every engineer gets the same enforcement on every file.
An MCP server and hooks ship together in one plugin, so policy runs where engineers already work.
The agent asks Semgrep and Semgrep answers. The MCP server exposes Semgrep scanning as tools the agent can call directly.
Learn more about the MCP →Hooks fire whenever Claude Code writes a file, ensuring a scan regardless of what the agent does or forgets.
Learn more about Hooks →Run your own Semgrep rules at generation time, in CI, or both. Put fast, deterministic checks in the agent session and leave heavier analysis to the pipeline.
Guardian reports what it catches to the Semgrep AppSec Platform, so security teams see findings and activity across projects in one place.
Roll out through the agent console or your MDM in a few hours.
Choose the plan that fits how you or your organization builds with AI.
Guardian-only for individual developers and teams of any size.
Bundles Guardian with Semgrep Code. Teams can still choose Guardian-only for $10 per seat.
For larger organizations planning a tailored rollout.
Bring Guardian into Claude Code so malicious packages, hardcoded secrets, and vulnerable patterns are caught before they reach a pull request.