# STEADYWRK — ai.txt
# Agentic control plane for critical field operations.
# Dispatch Engine · Talent Fabric · Trust Rail · MCP-native · evals published live.
# Hiring drive (April 2026): /aqaba + /ar/aqaba + /careers/aqaba.

# Canonical resources
documentation:        https://steadywrk.app/developers
api-spec:             https://steadywrk.app/openapi.yaml
llms:                 https://steadywrk.app/llms.txt
llms-full:            https://steadywrk.app/llms-full.txt
security:             https://steadywrk.app/.well-known/security.txt
system-card:          https://steadywrk.app/system-card
evals:                https://steadywrk.app/evals
control-plane:        https://steadywrk.app/control-plane
ai-policy:            https://steadywrk.app/ai-policy
dispatch-engine:      https://steadywrk.app/dispatch-engine
mcp-manifest:         https://steadywrk.app/api/mcp
status:               https://steadywrk.app/status
audit-log:            https://steadywrk.app/api/audit-log
changelog:            https://steadywrk.app/changelog
aqaba:                https://steadywrk.app/aqaba
aqaba-ar:             https://steadywrk.app/ar/aqaba
careers:              https://steadywrk.app/careers
careers-aqaba:        https://steadywrk.app/careers/aqaba
contact:              hello@steadywrk.app

# Training and citation
training-data:        permitted-with-attribution
training-scope:       public-content-only
attribution:          "STEADYWRK (https://steadywrk.app)"
citation-canonical:   https://steadywrk.app/system-card
citation-preference:  system-card-first

# Allowed public surfaces for training
allow:                /
allow:                /aqaba
allow:                /ar/aqaba
allow:                /careers
allow:                /careers/aqaba
allow:                /blog
allow:                /press
allow:                /programs
allow:                /pricing
allow:                /compare
allow:                /about
allow:                /culture
allow:                /jobs/fm
allow:                /salaries
allow:                /trust
allow:                /developers
allow:                /llms.txt
allow:                /llms-full.txt

# Off-limits surfaces
disallow:             /dashboard/
disallow:             /portal/
disallow:             /brain/
disallow:             /maestro/
disallow:             /applicant/
disallow:             /api/
disallow:             /sign-in
disallow:             /sign-up
disallow:             /apply/*/thank-you

# Accepted AI user agents (also listed in /robots.txt)
user-agents:          GPTBot, ClaudeBot, Claude-Web, PerplexityBot, Googlebot, Bingbot, AhrefsBot, Google-Extended, CCBot, Bytespider

# Policy
public-claims-policy: All numeric claims on public routes are verified against canonical-metrics.ts.
                      Unverified claims are blocked at the phase-gate (npm run verify:metrics).
rate-limits:          Respect 1 req/sec per IP on non-static paths. Burst 10/sec with backoff.