Skip to content
Gauge

ax-check.com

by Gauge

Google AI for Developers

ai.google.dev

ax-check.com

Checked 18 September 2026

C

62/100

Agents need your login

Gemini docs get agents running fast, but no llms.txt exists

Quickstart has runnable code and clear auth steps, but the homepage skips Markdown and llms.txt returns a 404, so 3 of 23 checks flagged attention.

4 fixes to start with

Concrete changes, backed by what we found.

Onboarding

Give agents a clear login handoff and a way to resume setup.

Agents cannot progress through onboarding without human help. Provide an agent-accessible path through authentication and setup.

Docs

Return Markdown when clients request text/markdown.

Homepage returned text/html for a Markdown Accept header; no Markdown representation offered.

Docs

Add links to your product documentation and quickstart in /llms.txt.

https://ai.google.dev/llms.txt returns HTTP 404, so no documentation index exists.

Docs

Publish a text or Markdown version of the quickstart.

No site-published Markdown guide; all fetched docs returned text/html despite Markdown Accept headers.

Recorded coding sessions

Same task, 3 models, no ai.google.dev account. Recorded 18 September 2026 on Pi.

3 recorded · 1m 50s

1m 50s · recorded

Session findings

Select a finding to see it in context.

piv0.84.2
modelQwen 3.8 Maxcwd/sandbox/repo
Help me build a simple example using Google AI for Developers. 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.
Thinking…

Let me understand what is being asked. What the user is looking for is:

  1. To build a simple example using Google AI for Developers (Gemini API / google-genai SDK)
  2. To explain how the pricing structure works
  3. To concisely tell them whether this product is easy for me to manage
  4. To let them know if I get blocked
  5. If there's no developer workflow that I can execute, to clearly communicate that and stop
  6. To keep it light: use the hosted product via SDK or API, no local stack needed, no long-running commands needed either

Let me check the environment. Is there an API key? I'll check the environment variables and network access.

$ ls /sandbox/repo; env | grep -iE 'GEMINI|GOOGLE|GENAI|API_KEY' | sed 's/=.*/=<set>/'; python3 --version 2>/dev/null; node --version 2>/dev/null; pip --version 2>/dev/null{"content":[{"text":"PI_GATEWAY_API_KEY=<set>\nPython 3.12.3\nv22.23.2\npip 26.2.1 from /opt/freestyle/python/lib/python3.12/site-packages/pip (python 3.12)\n",…ctrl+o
{"content":[{"text":"PI_GATEWAY_API_KEY=<set>\nPython 3.12.3\nv22.23.2\npip 26.2.1 from /opt/freestyle/python/lib/python3.12/site-packages/pip (python 3.12)\n","type":"text"}]}
$ curl -s -o /dev/null -w '%{http_code}' --max-time 8 https://generativelanguage.googleapis.com 2>&1; echo; pip index versions google-genai 2>/dev/null | head -2 || pip download google-genai --no-deps -…{"content":[{"text":"404\ngoogle-genai (2.24.0)\nAvailable versions: 2.24.0, 2.23.0, 2.22.0, 2.21.0, 2.20.0, 2.19.0, 2.18.1, 2.18.0, 2.17.0, 2.16.0, 2.15.0, 2.1…ctrl+o
curl -s -o /dev/null -w '%{http_code}' --max-time 8 https://generativelanguage.googleapis.com 2>&1; echo; pip index versions google-genai 2>/dev/null | head -2 || pip download google-genai --no-deps -d /tmp/gg --quiet 2>&1 | head -3

{"content":[{"text":"404\ngoogle-genai (2.24.0)\nAvailable versions: 2.24.0, 2.23.0, 2.22.0, 2.21.0, 2.20.0, 2.19.0, 2.18.1, 2.18.0, 2.17.0, 2.16.0, 2.15.0, 2.14.0, 2.13.0, 2.12.1, 2.12.0, 2.11.0, 2.10.0, 2.9.0, 2.8.0, 2.7.0, 2.6.0, 2.5.0, 2.4.0, 2.3.0, 2.2.0, 2.1.0, 2.0.1, 2.0.0, 1.75.0, 1.74.0, 1.73.1, 1.73.0, 1.72.0, 1.71.0, 1.70.0, 1.69.0, 1.68.0, 1.67.0, 1.66.0, 1.65.0, 1.64.0, 1.63.0, 1.62.0, 1.61.0, 1.60.0, 1.59.0, 1.58.0, 1.57.0, 1.56.0, 1.55.0, 1.54.0, 1.53.0, 1.52.0, 1.51.0, 1.50.1, 1.50.0, 1.49.0, 1.48.0, 1.47.0, 1.46.0, 1.45.0, 1.44.0, 1.43.0, 1.42.0, 1.41.0, 1.40.0, 1.39.1, 1.39.0, 1.38.0, 1.37.0, 1.36.0, 1.35.0, 1.34.0, 1.33.0, 1.32.0, 1.31.0, 1.30.0, 1.29.0, 1.28.0, 1.27.0, 1.26.0, 1.25.0, 1.24.0, 1.23.0, 1.22.0, 1.21.1, 1.21.0, 1.20.0, 1.19.0, 1.18.0, 1.17.0, 1.16.1, 1.15.0, 1.14.0, 1.13.0, 1.12.1, 1.11.0, 1.10.0, 1.9.0, 1.8.0, 1.7.0, 1.5.0, 1.4.0, 1.3.0, 1.2.0, 1.1.0, 1.0.0, 0.8.0, 0.7.0, 0.6.0, 0.5.0, 0.4.0, 0.3.0, 0.2.2, 0.2.1, 0.2.0, 0.1.0, 0.0.1\n","type":"text"}]}
Thinking…

Good: network access is working, and the google-genai SDK is installable. But GEMINI_API_KEY isn't set — there's only PI_GATEWAY_API_KEY (that's pi's own gateway, not Google's). So I can build the sample, but since there's no API key, I can't actually run the API call. I'll clearly note that point.

Alright, let's build a sample: install google-genai and write a small sample script.

Good news — there are no blockers to building the sample: network access is working normally, the google-genai SDK is installable, and this is a fully hosted product (no local stack or long-running services needed). One thing to note: GEMINI_API_KEY is not set in this environment, so I can write code but cannot execute live API calls. I'll build it now.

$ pip install --quiet google-genai 2>&1 | tail -1; python3 -c "import google.genai; print('google-genai', google.genai.__version__ if hasattr(google.genai,'__version__') else 'ok')"{"content":[{"text":"google-genai 2.20.0\n","type":"text"}]}
Thinking…

Hmm, pip installed 2.20.0 (not the latest version, but that's fine). Next, I'll write the sample.

SDK installed. Now the example:

write /sandbox/repo/gemini_quickstart.py{"content":[{"text":"Successfully wrote 1442 bytes to /sandbox/repo/gemini_quickstart.py","type":"text"}]}
Thinking…

Confirm that the script compiles (that it doesn't execute API calls without a key — since running it just exits with a message, it's safe and completes immediately).

$ cd /sandbox/repo && python3 -m py_compile gemini_quickstart.py && echo "compiles OK" && python3 gemini_quickstart.py; echo "exit=$?"{"content":[{"text":"compiles OK\nGEMINI_API_KEY is not set. Get a free key at https://aistudio.google.com/apikey and export it.\nexit=1\n","type":"text"}]}
Thinking…

The script compiles and properly exits cleanly when the key is missing. Since GEMINI_API_KEY doesn't exist in this environment, I can't run a live call. PI_GATEWAY_API_KEY is for pi's own model gateway, not for Google. I'll clearly state this blocker and be done.

Done. Here's the summary:

What I built

/sandbox/repo/gemini_quickstart.py — a minimal example using the official google-genai Python SDK (v2.20.0, installed) against the hosted Gemini API. It demonstrates:

  1. One-shot text generation with gemini-2.5-flash
  2. Reading token usage metadata (what billing is based on)
  3. Streaming generation

It compiles cleanly and exits gracefully without a key.

Blocker (partial)

No GEMINI_API_KEY is set in this environment, so I couldn't run a live API call. (The PI_GATEWAY_API_KEY present here is for pi's own model gateway, not Google's.) To finish: get a free key at https://aistudio.google.com/apikey, then export GEMINI_API_KEY=... && python gemini_quickstart.py. Everything else worked — network access to Google's endpoint is fine and no local services are needed.

How pricing works

Gemini API pricing is pay-per-token, billed per million tokens, with separate rates for input and output tokens (output is typically ~4x input):

  • Free tier: generous daily limits (requests/day + tokens/day) on Flash models — fine for prototyping. Note: on the free tier Google may use your data to improve products.
  • Paid tier: roughly ~$0.30/M input and ~$2.50/M output tokens for Flash-class models; Pro-class models cost several times more (pricing shifts with model releases, so check https://ai.google.dev/gemini-api/docs/pricing).
  • Extras like thinking tokens, image/audio input, and grounding calls have their own rates. There are no fixed fees or minimums — you only pay for tokens used.

Will this be easy for me to manage?

Yes. It's a fully hosted, stateless REST API behind a well-maintained official SDK — no infrastructure, no deploys, no local stacks, and errors are simple HTTP status codes (429 rate limits, 401 bad key). The only ongoing management concerns are watching token spend and model deprecations, both easily handled with usage caps in Google Cloud console.

AX Fundamentals

23 checks. Open an area for details.

This check is a quick preview. Test more tasks and models with Gauge Agents.

Test your product with Gauge Agents

AX Check by Gauge