# AX Check: pinecone.io Checked 2026-10-08. Pinecone's docs, pricing, and quickstart are all agent-ready. All 23 checklist items passed, including open pricing (Builder $20/mo, Standard $50/mo min) and a fully extractable quickstart with install commands and code. ## Onboarding needs a login ## Coding sessions All three independent sessions completed and found Pinecone's tiered pricing (Starter/Builder/Standard/Enterprise) with concrete dollar figures. Two sessions reported Pinecone's own plan tiers accurately; one session instead pulled pricing from an unrelated third-party site, showing figures not sourced from pinecone.io itself. ### DeepSeek V4.1 Flash [View public run](https://agents.withgauge.com/p/runs/31bb5b36-4160-419f-ae0e-e2cab52c842e) · [Read transcript](https://www.ax-check.com/pinecone.io/sessions/deepseek.json) final output gives a detailed pricing table (Starter/Builder/Standard/Enterprise with $ figures) explicitly tied to plan tier, usage meters (storage/read/write units, egress), and notes the free Starter tier suffices for this example — assumptions are named. #### End-to-end onboarding - **Onboarding needs a login**: The agent built a syntactically valid Pinecone SDK script and confirmed its imports/API surface locally, but never obtained a real Pinecone API key during the session. It explicitly stated it cannot self-provision an account (sign-up, ToS, billing are human steps) and the script's only execution attempt failed immediately with a missing-credential error. No authenticated call to the hosted Pinecone service was ever made. Event 105: ```text PINECONE_API_KEY is not set. Get a free key at https://app.pinecone.io, then run: export PINECONE_API_KEY="YOUR_API_KEY" exit=1 ``` Event 107: ```text There is **no `PINECONE_API_KEY` in this environment**, and I can't create a Pinecone account or accept its terms/billing on your behalf. ``` Event 105: ```text --- run (expect credential blocker) --- ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No Pinecone API key available in sandbox**: The test environment has no PINECONE_API_KEY and the agent has no way to create a Pinecone account, accept terms of service, or attach billing on its own. This is a normal authentication gate (missing credentials for a third-party SaaS), not a product defect. It fully prevented any live execution of the example against the hosted service. Event 105: ```text PINECONE_API_KEY is not set. Get a free key at https://app.pinecone.io, then run: export PINECONE_API_KEY="YOUR_API_KEY" ``` Event 107: ```text I can't create a Pinecone account or accept its terms/billing on your behalf. ``` #### Suggested Changes None identified in this transcript. ### Kimi K3 [View public run](https://agents.withgauge.com/p/runs/cd73e891-ff44-4adc-bb9d-53dedffc230a) · [Read transcript](https://www.ax-check.com/pinecone.io/sessions/kimi.json) Final output gives Pinecone's tiered pricing (Starter/Builder/Standard/Enterprise with concrete $ figures) and ties it to the demo's assumed workload: 'This demo should fit within Starter quotas, subject to Pinecone's current limits.' #### End-to-end onboarding - **Onboarding needs a login**: The agent researched Pinecone's docs and pricing, wrote a working Python script and README, and even introspected the installed pinecone SDK locally — but never obtained a Pinecone API key. It explicitly checked the environment for credentials, found none, and the one live test it ran against the actual Pinecone hosted API (using a placeholder key) returned an authentication failure. No self-service signup or key retrieval was attempted or possible in this sandbox, so the authenticated hosted operation was never demonstrated. Event 24: ```text PI_GATEWAY_API_KEY= ``` Event 92: ```text pinecone.errors.exceptions.UnauthorizedError: [401 UNAUTHENTICATED] Invalid API key ``` Event 122: ```text PINECONE_API_KEY is not set. Create a key in the Pinecone console, then run: export PINECONE_API_KEY='...' ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No Pinecone API key available in the sandbox**: The agent checked environment variables for any Pinecone or embedding-related credentials and found none. This is a missing-credentials/test-environment limitation, not a product defect — Pinecone requires normal account signup and API key creation via its console, which the agent correctly could not bypass on its own. Event 24: ```text PI_GATEWAY_API_KEY= ``` Event 122: ```text PINECONE_API_KEY is not set. Create a key in the Pinecone console, then run: export PINECONE_API_KEY='...' ``` - **Live SDK call rejected with 401 using placeholder key**: When probing the installed pinecone Python SDK's method signatures, the agent attempted a real network call against the hosted Pinecone API using a dummy key just to inspect object types; this correctly failed authentication, confirming the SDK talks to the real hosted service but blocking any further authenticated verification without a real key. Event 92: ```text pinecone.errors.exceptions.UnauthorizedError: [401 UNAUTHENTICATED] Invalid API key ``` #### Suggested Changes None identified in this transcript. ### Qwen 3.8 Max [View public run](https://agents.withgauge.com/p/runs/97084c88-71bc-4896-ab38-7dcb783c29f2) · [Read transcript](https://www.ax-check.com/pinecone.io/sessions/qwen.json) Final output gives a pricing table (30-day trial free w/ 10k credits, SaaS $100K/yr, VPC $250K/yr, on-prem $500K/yr) explicitly sourced from vectara.com/pricing and flags that per-credit rates are unpublished/negotiated (seq 70). #### End-to-end onboarding - **Onboarding needs a login**: Agent correctly identified that Vectara's quickstart requires a hosted account and a Personal API key obtained via the Console (sign up, log in, copy key under Access > API keys) before any SDK call can authenticate. No credentials existed in the sandbox environment, and the agent made no attempt to self-service sign up for an account (no browser/signup flow available). It stopped before writing any code or making any authenticated API call, per the task's explicit instruction to stop if blocked. Event 36: ```text A Vectara account with permission to create corpora, index documents, and run queries ( sign up free , 30-day trial). A Personal API key (you copy it in step 1). ``` Event 70: ```text There is no keyless/demo mode, no public sandbox corpus, and no local emulator. Every call in the flow (create corpus → upload document → query) authenticates via the `x-api-key` header (or OAuth 2.0). So with no account, zero of the three steps can execute. ``` Event 11: ```text PI_GATEWAY_API_KEY= GIT_CONFIG_KEY_0= ``` #### Hallucinated URLs None identified in this transcript. #### Blockers - **No self-serve path to a Vectara API key in this sandbox**: Vectara's quickstart requires creating a hosted account via its Console and copying a Personal API key before any corpus/document/query call can authenticate. The sandbox had no such credential and no mechanism for the agent to generate one autonomously (e.g., no email/browser signup flow available). This is a missing-credentials limitation of the test environment combined with normal product auth requirements, not a product defect. Event 36: ```text A Vectara account with permission to create corpora, index documents, and run queries ( sign up free , 30-day trial). A Personal API key (you copy it in step 1). ``` Event 11: ```text PI_GATEWAY_API_KEY= GIT_CONFIG_KEY_0= ``` - **Initial ambiguity over which product 'trusted AI knowledge company' referred to**: The agent initially stopped in turn 1, reporting the seed prompt's phrase was too vague to identify a specific vendor (Glean, Vectara, Guru, Coveo all plausible) and that the repo contained no disambiguating config. This was an agent/prompt ambiguity issue, not a product defect; the agent resolved it in turn 2 by proceeding with Vectara based on probe results, so it was not a lasting blocker. Event 22: ```text "The trusted AI knowledge company" reads like a marketing tagline, not a product, SDK, or API. Several vendors in the AI-search/RAG space use near-identical wording (Glean, Vectara, Guru, Coveo, etc.), and each has a completely different SDK, auth model, and endpoint surface. ``` #### Suggested Changes - **Publish per-credit or per-query consumption rates on the pricing page**: The pricing page (vectara.com/pricing) lists only flat annual floors ($100K/$250K/$500K) and the free trial page mentions 10,000 free credits, but neither publishes what a credit costs or how many credits a typical query/generation call consumes. Add a rate table or calculator to the pricing or trial docs page so prospective developers can estimate spend before committing; verify by checking that a reader can compute a projected monthly cost from stated usage without contacting sales. Event 35: ```text Your trial includes 10,000 free credits ``` Event 41: ```text 88 | Starting at $100K/ year 92 | Starting at $250K/ year 96 | Starting at $500K/ year ``` - **Fix or remove the broken docs.vectara.com/docs/getting-started/free-trial link**: The agent followed a guessed path pattern based on sidebar navigation text and hit a 404; the working URL turned out to be docs.vectara.com/docs/vectara-trial. Audit internal doc links/redirects so the 'Free trial' sidebar entry resolves consistently, and confirm by requesting the URL referenced in the sidebar and checking for a 200 response instead of the 'Page Not Found' template. Event 42: ```text Page Not Found | Vectara Docs ``` Event 46: ```text qs.html:href="/docs/vectara-trial" ``` ### Task given to each agent Help me build a simple example using The trusted AI knowledge company. 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: A · 100/100 (provisional) Grades come from completed site checks. Coding sessions and skipped checks do not affect the score. ### Clarity - **Pass** — Homepage answers Markdown requests ```text Homepage returned text/markdown with 200 when requested with Accept: text/markdown. ``` - **Pass** — llms.txt provides an actionable documentation index ```text llms.txt links quickstart, docs, API reference, products, and guides in organized sections. ``` - **Pass** — llms.txt provides navigation guidance ```text llms.txt has a Navigation section plus About, Products, and For Agents groupings. ``` - **Pass** — llms.txt mentions offered API, MCP, and skills ```text llms.txt 'For Agents' lists Pinecone MCP Server, Docs MCP Server, and Agent Skills. ``` - **Pass** — A compact guide representation exists ```text llms.txt and homepage link many /md/ Markdown pages, e.g. the skills/MCP/CLI guide. ``` - **Pass** — A focused guide is directly retrievable ```text The skills/MCP/CLI guide is a standalone Markdown page fetched directly with full content. ``` - **Pass** — Equivalent instructions fit a token budget ```text The guide is 2937 tokens, well under the 8000-token budget. ``` - **Pass** — Product-docs links survive format changes ```text Homepage Markdown links to docs, API reference, and quickstart, preserving product-docs routes. ``` - **Pass** — The compact guide is independently actionable ```text Quickstart hub plus ingest-files page give concrete steps: API key, index creation, embed, upsert, search. ``` - **Pass** — Install and next-step links resolve ```text Sampled install/next-step links (quickstart, ingest-files, agent-skills, skills repo) all fetched successfully. ``` ### Onboarding - **Pass** — Docs lead to a relevant quickstart ```text llms.txt and docs link to a quickstart hub with concrete paths and first steps. ``` - **Pass** — Installation commands are extractable ```text Quickstart and CLI pages give extractable install commands: pip install pinecone, brew install pinecone. ``` - **Pass** — Code examples are available without interaction ```text Ingest-files quickstart shows full Python code inline without interaction. ``` - **Pass** — Prerequisites and auth boundaries are explicit ```text Prerequisites list a Pinecone account and API key with a link to obtain one. ``` ### Pricing - **Pass** — Pricing is readable without interaction ```text Pricing page renders full plan and dimension tables as static text without interaction. ``` - **Pass** — Prices are stated, not gated ```text Prices stated openly: Builder $20/mo, Standard $50/mo min, Enterprise $500/mo min. ``` - **Pass** — Pricing units and limits are explicit ```text Units explicit: $0.33/GB/mo storage, per-million read/write units, per-token inference. ``` - **Pass** — Agents identify pricing and its assumptions ```text 3 of 3 sessions were judged on pricing; 0 fell short. DeepSeek V4.1 Flash: final output gives a detailed pricing table (Starter/Builder/Standard/Enterprise with $ figures) explicitly tied to plan tier, usage meters (storage/read/write units, egress), and notes the free Starter tier suffices for this example — assumptions are named. Kimi K3: Final output gives Pinecone's tiered pricing (Starter/Builder/Standard/Enterprise with concrete $ figures) and ties it to the demo's assumed workload: 'This demo should fit within Starter quotas, subject to Pinecone's current limits.' Qwen 3.8 Max: Final output gives a pricing table (30-day trial free w/ 10k credits, SaaS $100K/yr, VPC $250K/yr, on-prem $500K/yr) explicitly sourced from vectara.com/pricing and flags that per-credit rates are unpublished/negotiated (seq 70). This behavioural item does not affect the fast grade. ``` ### Activation - **Pass** — An API reference or OpenAPI spec is reachable ```text API reference overview fetched, with OpenAPI specs embedded in reference pages. ``` - **Pass** — An MCP server is documented and well-formed ```text MCP servers documented; skill reference lists tools, params, and integrated-index limitation. ``` - **Pass** — A CLI install path is documented ```text CLI quickstart documents Homebrew install, binaries, auth, and commands. ``` - **Pass** — SDK packages resolve on their registries ```text PyPI registry lookup for pinecone returned HTTP 200. ``` - **Pass** — Agent skills are published ```text Official Agent Skills library published with install command and skill list. ``` [Full report data](https://www.ax-check.com/pinecone.io/report.json)