API + CLI + MCP

Measure AI-written code where you already work.

Facet gives your coding agent an IP-safe repair plan, then compares the changed profile with its measured baseline.

One loop, three access surfaces

CLIRun Facet from a terminal or coding-agent workflow, with JSON output for automation.

MCPLet an MCP-capable coding agent request findings, repair plans, and comparisons in context.

APIIntegrate the same versioned contracts into CI, editors, internal tools, or your own product.

The surfaces use the same public contracts and the same account credit balance. You do not need to buy a separate CLI licence or a separate MCP plan.

Measure. Improve. Verify.

1. Measure

Profile the baseline revision. Capabilities and violations remain separate, so one cannot hide the other.

2. Improve

Request a violation-reduction plan, a capability-uplift plan, or both. Your agent sees the change objective and verification steps, not Facet's private indicators.

3. Verify

Remeasure the changed revision and compare it with the baseline. The result reports net violation reduction and net capability uplift. This is a finding comparison, not proof of functional correctness or complete security.

Self-service access

Sign in to Facet.
Create or revoke an access key. The raw key is shown once.
Run the CLI in your terminal, or add the local MCP server to Codex, Claude Code, or Cursor.

One Facet account, key, and credit balance works across the API, CLI, MCP, and website. Keys expire after 90 days and can be revoked without changing your sign-in.

Copyable setup

Create a key, replace the bracketed value below, then paste each line into PowerShell. The token stays in your user environment and is never placed in an agent prompt.

1. Install the Facet CLI and local MCP server

py -m pip install --upgrade facet-code

2. Connect the client to Facet

setx FACET_API_URL "https://api.facetcode.dev"
setx FACET_API_TOKEN "<KEY_FROM_THE_ACCESS_PAGE>"

3. Open a new terminal and check the connection

facet doctor --json

4. Add Facet to Codex

codex mcp add facet -- facet-mcp

Claude Code alternative

claude mcp add --transport stdio facet --scope user -- facet-mcp

Cursor alternative: paste into ~/.cursor/mcp.json

{
  "mcpServers": {
    "facet": {
      "command": "facet-mcp",
      "args": []
    }
  }
}

Cursor reads the global configuration from ~/.cursor/mcp.json. Keep tool approval enabled. After setup, ask your coding agent: Use Facet to map this repository and estimate a D6 security scan. Do not start paid work.

How beta charging works

A fresh before or after measurement uses profile credits.
A cached repair plan costs no additional profile credit.
CLI, MCP, API, and web access draw from the same account allowance.
Unchanged cached code costs zero to retrieve again.
There is no separate remediation surcharge during the beta.

Your first 10 full profiles are free during beta. When they are used, Stripe checkout is available for Pro or a 25-credit top-up. Pro adds monthly capacity and concrete fix generation. See current AUD pricing and credits.

What the beta does not automate

Facet does not apply the patch for you yet. It produces the IP-safe repair plan; you or your coding agent changes the code; Facet then evaluates the before-and-after findings. Fully orchestrated patch application and signed verification receipts remain roadmap work.

This public beta tests whether the loop is useful inside real AI-assisted coding sessions. Facet does not call it autonomous remediation or functional verification.

Where Facet runs

The website and API run on Facet’s managed infrastructure. You do not need a virtual machine or an always-on computer.

Repository mapping, Git comparison, and the MCP bridge run locally while you use them. A full profile sends only the code you explicitly submit to the hosted measurement API. Read the source-handling boundary.

Follow the end-to-end testing guide