I scanned 14 MCP servers on my laptop, then compiled a least-privilege policy from what my agent actually did

Description and use

The author discovered 14 MCP servers and 5 plaintext API keys on a laptop, then used pod to generate a least‑privilege policy from observed agent behavior and created a cryptographically verifiable audit trail.

Problem

Multiple agent platforms on a single laptop exposed 14 MCP servers, 12 unpinned, and 5 plaintext API keys in config files, creating a security risk due to unknown permissions and exposed credentials.

Outcome

The author successfully identified 5 plaintext API keys and 12 unpinned MCP servers, then generated a least‑privilege policy that denies unnecessary file access (e.g., read_file on .env) and only grants permissions observed in the recorded corpus. The audit trail is cryptographically verifiable, ensuring transparency and ownership of the policy.

Workflow

• 1. Install pod on macOS/Linux via the provided install script.
• 2. Run pod in record‑only mode to capture agent behavior for a few days.
• 3. Compile a policy draft from the recorded calls using pod’s compile step.
• 4. Serve the policy with pod serve, enforcing deny‑first, approve, allow, and fail‑closed logic.
• 5. Verify the audit trail with pod verify‑audit, producing a hash chain that can be handed to others for independent verification.

Evidence

I run six agent platforms on one laptop. Between them they have 14 MCP servers configured, 12 of them unpinned (npx -y pkg@latest), and five plaintext API keys sitting in config files.

Source

record run your agent through pod in record-only mode for a few days compile pod policy draft -> least-privilege rules from real observed calls enforce pod serve -> deny > approve > allow, fail-closed prove pod verify-audit -> SHA-256 hash chain you can hand to someone else

Source