›Spectra_cli
⌘K Install
// built for engineers shipping with agents

The readiness check
for the agent era.

One command scores how well an AI agent can work in your repo, shows you the gaps with evidence, and fixes them in a tight, verifiable loop. Runs locally. Bring your own model.

$npx spectra scan .
no code uploaded scan in seconds any git repo
~/work/payments-api
$ spectra scan .
// the loop that makes agents trustworthy

Make a change. Verify. Correct. Repeat.

Spectra optimises for the thing that lets an agent work unsupervised: a fast, machine-runnable feedback loop it can trust.

01

Detect

Repo type, toolchain and the checks an agent can actually run.

02

Verify

typecheck, lint and test — run on demand, results trusted.

03

Fix

Agents work in worktrees with diff previews and a kill switch.

// one tool, the whole workflow

From scan to merged, on the command line.

$ spectra scan .→ assessing six readiness lenses against ./payments-api✔ agent context  35  ·  no AGENTS.md at repo root→ readiness 61/100 · 5 high-impact actions
// real output, not a marketing mock

Findings you can act on.

spectra report --top 

              FAIL
              tribal knowledge
              ▰▰▰▰▰▱▱▱▱▱
              48
              timeout = 3000 // don't change — no rationale recorded
            
              PASS
              verifiability
              ▰▰▰▰▰▰▰▰▱▱
              82
              typecheck script present · 0 errors · suite runs in 11s
            
              WARN
              reproducibility
              ▰▰▰▰▰▰▱▱▱▱
              64
              No pinned runtime (.nvmrc / .tool-versions) detected
            
              PASS
              structure
              ▰▰▰▰▰▰▰▱▱▱
              71
              2 files over 1,000 lines · likely load-bearing
            
              FAIL
              agent context
              ▰▰▰▰▱▱▱▱▱▱
              35
              No AGENTS.md or .cursor/rules found at repo root
            
              WARN
              safety
              ▰▰▰▰▰▰▱▱▱▱
              58
              Test suite contains a live HTTP client — unguarded
            ────────  overall 61/100 · material gaps · run `spectra plan`
// six lenses

What “ready” actually measures.

Get your readiness score now.

$npx spectra scan .
 or Download the desktop app

Free for solo use · macOS, Windows, Linux · bring your own model key