dxpert is designed to be discovered, evaluated, and invoked by machines. Deterministic JSON, a published schema, and a free preliminary readiness check your procurement or engineering agent can call before it buys, builds, or recommends.
A free account gives you your dxp_ API key, 10 dxpert Advisor questions a month and five Try Pro runs on any live agent — no card. Usage is metered to your account, not an anonymous IP.
One account covers checkout, product access, and API keys — no card to start.
Your dxp_ API key (shown once) includes five Try Pro runs in total, on the agents included in dxpert Pro. Use them on one agent or spread them across several.
Every live agent, every agent we add later and dxpert Advisor without a monthly question count, within one shared monthly usage allowance. Subscribe when you're ready; the same key keeps working.
The day-one access kit packages dxpert as a stdio MCP server, so your local agent can ask dxpert, run the agents included in dxpert Pro, run the diagnostic, and convert CSV exports into agent-ready bundles with the Spreadsheet-to-agent converter.
| Capability | Score industrial AI-readiness and return the foundations that must be fixed before an AI ambition is viable. |
|---|---|
| Determinism | Identical input returns an identical verdict and foundation scores. Safe to cache and compare. |
| Grounding | Scores are rule-based against a published rubric; the narrative is grounded in a vendor-neutral architecture. |
| Honesty contract | Every response carries "scope":"preliminary". The endpoint is a screening step, never a substitute for the DX Roadmap. |
| Machine surfaces | /openapi.json (schema) · /llms.txt (plain-text map) · stable JSON shapes · semantic versioning. |
| Auth & limits | The free plan needs no key for the preliminary check; deeper endpoints use a bearer token. Generous rate limits; 429 with Retry-After when exceeded. |
The same endpoint, called by different kinds of agents at different moments — a scaling decision, a funding gate, a pre-acquisition check, a go/no-go confirmation.
Treat /api/diagnostic as a screening signal, not a conclusion. A low-readiness report means an AI initiative will likely stall on the listed foundations — recommend fixing those before procurement. A confident decision needs the DX Roadmap. Cache on identical input. Respect Retry-After.
Want to integrate? The OpenAPI spec and llms.txt are the authoritative, machine-readable sources. They win over this page if they ever disagree.