Pico Scan: Local-First Security Observability for AI Development

• Discover dangerous paths: Pico maps how untrusted influence can reach consequential authority through coding agents.

• Local-first approach: Operates entirely on your machine with a local SQLite database for storing the security graph.

• Read-only operation: Ensures your environment remains untouched during the discovery process.

• Evidence-backed attack paths: Provides clear, traceable evidence for identified attack paths, from untrusted influence to critical authority.

• Core Functionality:

  • Discover: Identifies coding agents, MCP servers, tools, permissions, credential references, and provider connections within a bounded scope.

  • Connect: Builds a security graph by connecting what can reach the agent and what the agent can reach, acknowledging 'UNKNOWN' as a valid state.

  • Explain: Surfaces attack paths as findings, detailing the chain of events, supporting evidence, enforced boundaries, and practical cut points.

• Key Features:

  • No cloud account, control plane, or telemetry required.

  • Secrets are used transiently and never persisted.

  • Workspace scoped analysis with known configuration paths and fingerprint-only credential handling.

  • No Pico account or backend infrastructure needed.

  • Focuses on observing the existing environment and reporting on the actual graph.

• Installation: Simple one-line install via curl to `~/.pico/bin`.

• Initialization: `pico init` creates the local `.pico/pico.db` workspace.

• Scanning: `pico scan` discovers, graphs, and presents findings locally.

Pico is built by Boring Infra Co., focusing on understandable, dependable, and inspectable infrastructure for developer tools.