Skip to content

START_HERE.md

Latest commit

 

History

History
177 lines (130 loc) · 4.21 KB

START_HERE.md

File metadata and controls

177 lines (130 loc) · 4.21 KB
Raw

START_HERE.md — Where to begin with AMC

AMC can look broad because it is broad. This page is the shortest route to the right starting point.

What AMC is

AMC is a trust score for AI agents.

If you want the repo-backed architectural version of that statement before choosing a path, read docs/ARCHITECTURE_BRIEF.md.

It helps you:

  1. Score an agent from evidence
  2. Find trust and governance gaps
  3. Generate fixes, reports, and next actions

If you only remember one thing, remember this:

AMC is strongest when you want evidence, not vibes.

Choose your path

Path 1 — I just want to try it right now

Use the browser playground.

  • URL: website/playground.html
  • Best for: first-touch demos, lightweight exploration, understanding the scoring model
  • Limitation: browser try-now is for exploration, not full execution evidence capture

Next step:

  • Go to the playground
  • Explore the questions and scenarios
  • If you want real traces / datasets / CI gates, move to the CLI path

Path 2 — I want a real score locally

Use the CLI.

curl -fsSL https://agentmaturity.co/install.sh | sh
amc

On Windows PowerShell, install with irm https://agentmaturity.co/install.ps1 | iex, then run amc.

The first run is an honest baseline. A signed VALID artifact can still be INSUFFICIENT_EVIDENCE; capture a real agent run and rerun before making external claims. Only evidence readiness READY is claim-eligible.

Best for:

  • scoring a real project
  • getting a trust maturity level
  • seeing practical gaps
  • generating fixes

Recommended next steps after your first score:

  1. amc
  2. amc fix
  3. amc doctor --json
  4. review docs/AFTER_FIRST_SCORE.md

Path 3 — I want AMC in CI

Use the GitHub Action / CI workflow.

Best for:

  • score thresholds
  • preventing trust regressions
  • PR comments and artifacts
  • repeatable release gates

Start here:

  • .github/workflows/amc-score.yml
  • docs/CI_TEMPLATES.md

Path 4 — I want compliance and governance outputs

Use AMC when you need more than a score.

Best for:

  • EU AI Act mapping
  • audit binders
  • governance evidence
  • regulated delivery workflows

Start with:

  • Comply concepts in the README
  • compliance docs in docs/
  • binder/report generation workflows

Path 5 — I already have an agent stack

AMC works best when it wraps what you already run.

Examples:

  • LangChain
  • CrewAI
  • AutoGen
  • OpenAI Agents SDK
  • Claude Code
  • Gemini
  • OpenClaw
  • generic CLI agents

Use:

amc wrap <adapter> -- <your command>

Then move into:

  • full scoring with amc
  • traces
  • assurance packs
  • CI

AMC product family

These names are canonical:

  • Score — trust scoring and maturity diagnostics
  • Shield — adversarial assurance packs
  • Enforce — policy controls and approvals
  • Vault — signatures, proof chains, evidence integrity
  • Watch — traces, anomalies, monitoring
  • Fleet — multi-agent oversight and inventory
  • Passport — portable identity and credential artifacts
  • Comply — compliance mapping and audit outputs

Do not overthink this on day one. If you are new, start with:

  • Score
  • then Shield
  • then Watch / Comply as needed

Recommended first 15 minutes

  1. Install from the verified release script, then run amc
  2. Read the gaps
  3. Run amc fix
  4. Read docs/AFTER_FIRST_SCORE.md
  5. Decide whether you need:
    • browser exploration
    • local CLI workflows
    • CI gating
    • compliance outputs

Honest scope notes

  • The browser path is a real playground, not a fake full browser execution runtime.
  • The CLI is the serious path for execution evidence, traces, datasets, and CI.
  • Single-binary packaging exists as an experimental path and should be treated honestly.
  • SDKs, editor integrations, and CI shells are useful, but most of them wrap the same TypeScript runtime rather than replacing it.
  • If you want that distinction spelled out, read docs/IMPLEMENTATION_REALITY_MAP.md.

Next docs to read

  • docs/ARCHITECTURE_BRIEF.md
  • docs/IMPLEMENTATION_REALITY_MAP.md
  • docs/deep-dive/INDEX.md
  • docs/AFTER_FIRST_SCORE.md
  • docs/QUICKSTART.md
  • docs/ADAPTERS.md
  • docs/CI_TEMPLATES.md
  • docs/BROWSER_SANDBOX.md
  • docs/SINGLE_BINARY.md