AI-Assisted Workflow

Axis is built to be legible to AI coding assistants as well as humans. Three surfaces make an assistant productive without leaking secrets: the axis.json manifest, axs context, and axs doctor.

axs context — a redacted project digest

axs context writes a redacted digest — the directory tree, axis.json, the feature toggles, and the lambda/table registries — for pasting into an external AI assistant. It is static and offline: no network call, no AI round-trip, no synth.

axs context                 # print a redacted digest to stdout
axs context --out ctx.txt   # write to a file
axs context --copy          # copy to the OS clipboard (best-effort)
axs context --json          # structured JSON (digest + redactionReport)

Redaction runs in three layers: secret-bearing files are never read (config/*-config.ts, .env*, .npmrc, keys); secret-shaped values are scrubbed (AWS keys, account ids in ARNs, Bearer/JWT, KEY/SECRET/TOKEN assignments, emails); and a project-root .axisredact adds project-specific rules. The stderr summary lists exactly what was excluded — review before pasting; redaction is best-effort, not a guarantee.

axs doctor — a machine-readable health check

axs doctor is a static, read-only project health check — manifest coherence, registration drift, and auth — that exits non-zero on findings, so it gates a CI step. --json emits a stable findings array for CI or AI piping.

axs doctor
axs doctor --json

The manifest is the contract

axis.json is the machine-readable placement spec assistants read to put generated code in the right spot. See The axis.json Manifest.

Where to go next

• axs context and axs doctor references.
• The axis.json Manifest for the contract assistants parse.