Code Security

Operating model

What runs locally

Static analysis runs locally. Dependency inventories and lookups may use external services; AI features send requested code context to the configured provider.

See the VS Code security page Browse use cases

Why local-first matters

What teams are really asking when they ask for local-first security

What teams are really asking when they ask for local-first security
The questions behind the phrase How Oryon draws the line
Will developers get useful feedback fast enough to act while they are still coding? Scan as you edit or save, or start a file or repository scan; configure profiles and exclusions.
Does the operating model require sending full source code away just to get the first security review? Static analysis runs locally. Dependency inventories and lookups may use external services; AI features send requested code context to the configured provider.
Can the team preserve privacy and speed without losing shared visibility later? Dashboard sync exists for team memory, history, and shared state once the repository is linked.

How it works

Where analysis stays local and where team state becomes shared

  1. Run the core scan locally

    What happens

    Static analysis of code, secrets, infrastructure, pipelines and AI/MCP security. Combines patterns with data tracking within each file. Profiles and exclusions adapt the scan to the repository.

    What it adds

    That keeps the first security loop aligned with developer speed instead of waiting for a remote pass.

  2. Use AI after the scan, not instead of it

    What happens

    Explains causes, risk and impact and proposes changes, diffs, effort and verification checks. Available according to plan and configuration; the user reviews and applies the proposal in the editor.

    What it adds

    That boundary makes the product easier to reason about for privacy, speed, and operational trust.

  3. Sync team memory only when the repo is linked

    What happens

    Linked repositories can sync findings, dependency vulnerabilities, suppressions, and scan history to the dashboard.

    What it adds

    That gives the team shared visibility and continuity without moving the analysis engine itself out of the editor workflow.

Best fit

When local-first is the right operating choice

Choose Oryon if

  • Privacy, speed, and developer adoption all matter in the same decision.
  • You want the editor to be the first place security feedback appears.
  • You still need shared history and dashboard visibility once the repository becomes team-wide.

Choose another model if

  • You are comfortable with a cloud-first scanning model as the primary operating layer.
  • Developer latency is secondary to centralizing everything in one platform surface.
  • The team is not trying to preserve a strong local workflow at all.

FAQ

Questions teams ask when evaluating local-first security

Does local-first mean nothing is ever synced?

No. It means the core scan and first feedback loop stay local. Team memory can still be synced later when the repository is linked. Static analysis runs locally. Dependency inventories and lookups may use external services; AI features send requested code context to the configured provider.

Does AI require uploading the full codebase?

Static analysis runs locally. Dependency inventories and lookups may use external services; AI features send requested code context to the configured provider.

Why does local-first matter for adoption?

Because the first signal arrives faster, inside the existing editor workflow, and without forcing developers to depend on a separate remote review loop for every change.

Local-first only matters if it makes the workflow faster, clearer, and easier to trust.

The operating model should help the team move earlier without giving up the shared state needed after the first scan.