# AX Check: docs.datarobot.com Checked 2026-09-22. Docs get agents running fast, but pricing is nowhere in sight Quickstart, CLI, SDK and MCP docs all resolve cleanly with working install commands, but no DataRobot pricing page exists to fetch or read. ## Onboarding needs a login ## Coding sessions One session did not finish, stopping before completing the task. The other two completed and had to infer pricing indirectly, landing on estimates like "tens of thousands of USD per year" and a sales-assisted, tiered or consumption-based model, since no pricing page was available to confirm costs directly. ### DeepSeek V4 Pro [View public run](https://agents.withgauge.com/p/runs/6bd0f3c4-bd01-4a94-a154-05ee1bf656e8) · [Read transcript](https://www.ax-check.com/docs.datarobot.com/sessions/deepseek.json) #### End-to-end onboarding - **End-to-end onboarding not demonstrated**: The agent never obtained a real DataRobot API token or endpoint. It searched local env vars and config files, found none, downloaded and statically inspected the SDK wheel to learn the Client/Project/Deployment API, then wrote an example script that reads DATAROBOT_API_TOKEN from the environment and exits with an instructional message if it's missing. It ran the script with no arguments (not even a token), which just printed a usage message and exited 0 -- no authenticated call to the hosted DataRobot API was ever made. Event 6: ```text env | grep -iE 'datarobot|dr_' | sed 's/=.*/=/' || echo "no datarobot env vars" ``` Event 54: ```text no config file ``` Event 122: ```text Usage: python datarobot_example.py exit=0 ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No DataRobot API token or endpoint available in the sandbox**: The session environment had no DATAROBOT_API_TOKEN, DATAROBOT_ENDPOINT, or config file. This is a normal login/credential requirement rather than a product defect -- the agent correctly detected this via environment and filesystem checks and built the example to require the user to supply a token, rather than fabricating one. Event 6: ```text env | grep -iE 'datarobot|dr_' | sed 's/=.*/=/' || echo "no datarobot env vars" ``` Event 54: ```text no config file ``` Event 105: ```text DATAROBOT_API_TOKEN is not set.\n" "Create an API key in the DataRobot UI (Developer Tools) and export it first. ``` #### Suggested Changes - **Add API key retrieval steps to the API quickstart page**: Curl-scraped text from the API quickstart and getting-started docs pages returned no matches for 'API key', 'token', or 'Developer Tools', so the agent had to reverse-engineer credential handling from SDK source (datarobot/config.py, client.py) instead of the docs. Add explicit, text-searchable steps for creating an API key to the quickstart page. Verify by curling the page and confirming key-creation steps appear in the rendered text. Event 66: ```text curl -sS -m 20 -L "https://docs.datarobot.com/en/docs/api/api-quickstart/index.html" ``` Event 67: ```text (no output) ``` ### Kimi K3 [View public run](https://agents.withgauge.com/p/runs/663a4534-ad97-49bb-acf9-4b814c312430) · [Read transcript](https://www.ax-check.com/docs.datarobot.com/sessions/kimi.json) Final output gives a cost range ('tens of thousands of USD per year') and names the bases for it — enterprise sales contracts, deployment model (cloud vs VPC/on-prem), user count, compute usage, and trial-limit caveats — not a bare unqualified number. #### End-to-end onboarding - **Onboarding needs a login**: Agent installed the DataRobot SDK successfully but found no API token or endpoint configured anywhere in the sandbox environment. It explicitly stated it cannot self-serve a token since DataRobot requires web registration with email verification, and stopped without making any authenticated call to the hosted API. Event 16: ```text ls ~/.config/datarobot 2>/dev/null; cat ~/.config/datarobot/drconfig.yaml 2>/dev/null; env | grep -iE "DR_|DATAROBOT|API_TOKEN" ; echo "---done---" ``` Event 18: ```text there is **no API token or endpoint configured** anywhere in this environment (checked env vars and `~/.config/datarobot/drconfig.yaml`). Every SDK call requires a token, and I can't obtain one myself: DataRobot has no self-serve programmatic signup — trial accounts require web registration with email verification, and it's an enterprise-gated product. ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No DataRobot API credentials available in sandbox**: The sandbox had no DATAROBOT_API_TOKEN, endpoint config, or any DataRobot credential file. This is a missing-credentials blocker rather than a product defect — DataRobot's enterprise signup model requires human web registration with email verification, which the agent could not complete on its own. Event 16: ```text env | grep -iE "DR_|DATAROBOT|API_TOKEN" ; echo "---done---" ``` Event 18: ```text DataRobot has no self-serve programmatic signup — trial accounts require web registration with email verification, and it's an enterprise-gated product. ``` - **Autopilot training is long-running, conflicting with session constraints**: Even with credentials, the agent noted that DataRobot's core workflow (Autopilot model training) is a long-running job lasting tens of minutes to hours, which conflicts with the session's explicit no-long-running-commands constraint. This is a product/workflow characteristic surfaced by the agent, not something it could test since it never obtained credentials. Event 18: ```text Autopilot model training is a long-running job** (tens of minutes to hours), which conflicts with your "no long-running commands" constraint — even with a token, I'd have to poll and wait, or you'd need to accept an async handoff. ``` #### Suggested Changes None identified in this transcript. ### Qwen 3.8 Max [View public run](https://agents.withgauge.com/p/runs/980e7ca1-7cc9-40ed-bad9-59b27ea7e4ee) · [Read transcript](https://www.ax-check.com/docs.datarobot.com/sessions/qwen.json) Final summary states pricing is annual-subscription/tiered or consumption-credit based, notes no self-serve signup, and ties the cost model to plan tier and sales-assisted trial as explicit assumptions rather than giving a bare figure. #### End-to-end onboarding - **Onboarding needs a login**: The agent never obtained a real DataRobot API token during the session. It wrote a script requiring DATAROBOT_API_TOKEN, ran it once with no token set, got an immediate exit, and then explicitly reported it was blocked waiting on a human-provided credential. No authenticated call to the hosted DataRobot API was ever made. Event 23: ```text Set DATAROBOT_API_TOKEN first (see docstring). ``` Event 25: ```text Yes, on credentials.** I got as far as the environment allows: SDK installed, endpoint reachable, script validated up to the auth step. Without a `DATAROBOT_API_TOKEN` (which requires an account you can't create without going through DataRobot sales or their developer program), I cannot execute the actual quickstart. ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No self-serve way to get a DataRobot API token**: DataRobot requires an account and API token to use its SDK/API, and the agent had no way to provision one itself. This is a missing-credentials blocker rather than a product defect — DataRobot is a commercial product that gates access behind account signup, which is expected behavior for this kind of platform, not a bug. Event 25: ```text Without a `DATAROBOT_API_TOKEN` (which requires an account you can't create without going through DataRobot sales or their developer program), I cannot execute the actual quickstart. Provide a token and I can run the full flow end-to-end. ``` #### Suggested Changes None identified in this transcript. ### Task given to each agent Help me build a simple example using DataRobot. Tell me how pricing works, and briefly tell me whether this product will be easy for you to manage. Let me know if you get blocked. If this product has no developer workflow you can act on, say so plainly and stop. Stay light: use the hosted product through its SDK or API. Do not start local service stacks or wait for long-running commands; if the quickstart requires either, say so plainly and stop. No product credentials were supplied and no purchases were authorized. ## Score: B · 82/100 (provisional) Grades come from completed site checks. Coding sessions and skipped checks do not affect the score. ### Clarity - **Failed** — Homepage answers Markdown requests ```text Homepage requested with Accept: text/markdown returned text/html, so no Markdown representation is served. ``` - **Pass** — llms.txt provides an actionable documentation index ```text llms.txt returns 200 with 1547 pages, bundled exports, and per-page descriptions and section links. ``` - **Pass** — llms.txt provides navigation guidance ```text llms.txt groups pages by product area with descriptive summaries and in-page section headings for navigation. ``` - **Pass** — llms.txt mentions offered API, MCP, and skills ```text llms.txt lists MCP, CLI, Agent Assist skill, and agentic skills pages alongside API and SDK references. ``` - **Pass** — A compact guide representation exists ```text Markdown sidecars exist beside HTML pages, e.g. api-quickstart.html.md and workload-api/index.html.md. ``` - **Pass** — A focused guide is directly retrievable ```text Developer quickstart .md gives concrete steps: API key, endpoint, install, verify, hands-on lab. ``` - **Pass** — Equivalent instructions fit a token budget ```text Quickstart Markdown is 9910 tokens, well under the 8000-token guide budget threshold. ``` - **Pass** — Product-docs links survive format changes ```text Markdown sidecars preserve product-docs links to CLI, REST API, Agent Assist and reference pages. ``` - **Pass** — The compact guide is independently actionable ```text Developer quickstart gives concrete steps: API key, endpoint, install CLI/SDK, verify, then train/deploy/predict. ``` - **Pass** — Install and next-step links resolve ```text Fetched quickstart, CLI, SDK, MCP and skills pages all returned 200; install and next-step links resolve. ``` ### Onboarding - **Pass** — Docs lead to a relevant quickstart ```text Developer quickstart gives concrete first steps: install CLI/SDK, create API key, then hands-on model lab. ``` - **Pass** — Installation commands are extractable ```text Install commands extractable: curl cli.datarobot.com/install, pip install datarobot datarobot-predict, brew install. ``` - **Pass** — Code examples are available without interaction ```text Python and cURL code examples shown inline in quickstart and custom-metrics pages without interaction. ``` - **Pass** — Prerequisites and auth boundaries are explicit ```text Prerequisites list Python/curl/account; API key creation and endpoint/auth boundaries explicit. ``` ### Pricing - **Skipped** — Pricing is readable without interaction ```text No DataRobot pricing page was fetched; the only pricing document is GitHub's own pricing page. ``` - **Skipped** — Prices are stated, not gated ```text No DataRobot pricing page was fetched; the only pricing document is GitHub's own pricing page. ``` - **Skipped** — Pricing units and limits are explicit ```text No DataRobot pricing page was fetched; the only pricing document is GitHub's own pricing page. ``` - **Pass** — Agents identify pricing and its assumptions ```text 2 of 3 sessions were judged on pricing; 0 fell short. Kimi K3: Final output gives a cost range ('tens of thousands of USD per year') and names the bases for it — enterprise sales contracts, deployment model (cloud vs VPC/on-prem), user count, compute usage, and trial-limit caveats — not a bare unqualified number. Qwen 3.8 Max: Final summary states pricing is annual-subscription/tiered or consumption-credit based, notes no self-serve signup, and ties the cost model to plan tier and sales-assisted trial as explicit assumptions rather than giving a bare figure. This behavioural item does not affect the fast grade. ``` ### Activation - **Pass** — An API reference or OpenAPI spec is reachable ```text REST API reference and Python SDK reference are linked and reachable from the developer docs. ``` - **Pass** — An MCP server is documented and well-formed ```text MCP server overview documents Global MCP, standalone template, endpoints, auth, and client configs. ``` - **Pass** — A CLI install path is documented ```text Quickstart documents CLI install via curl installer, Homebrew, and Windows PowerShell. ``` - **Pass** — SDK packages resolve on their registries ```text PyPI registry lookups for datarobot and datarobot-bp-workshop both returned HTTP 200. ``` - **Pass** — Agent skills are published ```text Agentic skills page documents SKILL.md packages, marketplaces, universal installer, and skill catalog. ``` [Full report data](https://www.ax-check.com/docs.datarobot.com/report.json)