Weaviate
weaviate.ioWeaviate is a vector database for searching and storing data by meaning.
Can an AI agent use Weaviate? Yes, without signing up. Our agent got into Weaviate on its own in 3 minutes and did all 10 of its core tasks.
From finding it to leaving
What an agent managed without a person, at each step.
- On its own
Find
Docs it can act on: llms.txt, an OpenAPI and an MCP server.
485 doc pages read
- On its own
Get in
Runs locally, sign-up in a browser.
3 minutes
- On its own
Use
Did: create collections and store objects, with the server making the embeddings; search by meaning, by keyword, by both, and with filters.
10 of 10 core tasks
- Reaches the checkout
Pay
Stopped at the $45 a month charge.
- On its own
Leave
Nothing to cancel, deletes its account.
Getting in
2 of 6 ways in worked without a person.
The way that worked
Runs locally, no account3 minutes
- An inboxNot needed
- A browserNot needed
- A personNot needed
- Agent on a serverAPI, MCP, CLINoStops at sign-up: it works only in a browser.
- Assistant in its own browserLike ChatGPT agentPartlySigns up, confirms its email and does all of the core work. Payment: reaches the checkout, stopped at the $45 a month charge.
- Signed agentKnown to CloudflareYesThe site opens.
Every way in
| Way in | What happened | Why |
|---|---|---|
| Runs locally, no account | Worked | |
| Email and code | Worked | |
| Continue with GitHub | Not tried | Not part of this drive |
| Continue with Google | Not tried | Not part of this drive |
| Added by an admin | Not tried | Not part of this drive |
| A key from a person | Not tried | Needs a person's account |
What an agent can do with Weaviate
All 10 core tasks done, and 10 of 12 others. One task not tried isn't counted.
Core tasks
- done there
| Task | Result | Through |
|---|---|---|
| Create a collection (schema) | Done | API: done there |
| Insert or upsert an object into a collection | Done | API: done thereMCP: done there |
| Run a vector (near_text/near_vector) search | Done | API: done there |
| Run a BM25 keyword search | Done | API: done there |
| Run a hybrid search combining vector and keyword | Done | API: done thereMCP: done there |
| Query records with filters and sorting | Done | API: done thereMCP: done there |
| Ask a natural-language question answered from stored data (Query Agent Ask mode) | Done | API: done thereMCP: done there |
| Run retrieval-focused search across collections with routing (Query Agent Search mode) | Done | API: done thereMCP: done there |
| Store a memory for an AI agent (text, conversation, or pre-extracted facts) | Done | API: done there |
| Search stored agent memories (vector/BM25/hybrid/fetch) | Done | API: done there |
Reliability
85of 100
Core tasks that worked on all 3 runs
Handles mistakes
76of 100
Clear errors, retries without duplicates, honest limits
Surface tested
73of 100
Documented routes, tools and commands run and checked
Docs accuracy
84of 100
Documented operations that behaved as the docs say
Docs completeness
96of 100
How fully the docs describe each operation
What it can't do, and why
- Sign up and get the first key without a browser Sign-up and the first key are made in the web console.
- Use GraphQL on a new free cluster The cluster answers that it is switched off.
- Generate text with a model Needs a model provider key we did not have.
- Back up a cloud cluster The free cluster's admin key is refused.
- Buy a paid plan Only through a card form; no billing API or agent payment.
10 of 12 other tasks done
- Create text or multimodal embeddings with a hosted modelDone
- Isolate data per tenant (multi-tenancy)Done
- Back up and restore a collection or instanceDone
- Create a role and assign fine-grained RBAC permissionsDone
- Create and manage database usersDone
- Delete an objectDone
- Add or update a collection's schema fieldsDone
- List collections and read their schemaDone
- Delete the signed-in agent's memory by IDDone
- Check an Engram pipeline run's statusDone
- Choose a deployment type (Shared Cloud, Dedicated Cloud, self-hosted)Done
- Monitor the instance via Prometheus metricsPaid plan
- Generate text grounded in retrieved objects (RAG): Needs a model provider key we did not have.Blocked
- Deploy Weaviate self-hosted via Docker/KubernetesNot tried
The plans they need
- Back up and restorepaid cluster
- Prometheus metricspaid / self-hosted
Paid-plan tasks don't count toward the level. The drive used the free plan.
Surface audit: 169 documented operations, 141 run, 116 passed, 22 not as documented, 17 blocked (sandbox or plan limits), 12 left out to stay safe.
How an agent connects
MCP, API, OpenAPI, SDK and llms.txt found.
MCP server
All testedOfficial server. Cluster MCP 4 tools; Query Agent MCP 2 tools.
6 tools across its docs
API
110 testedREST, 125 operations, plus gRPC; Engram REST 6 operations.
131 operations across its docs
OpenAPI
YesSwagger 2.0, info.version 1.40.1.
CLI
None foundSDK
22 testedweaviate-client 3.15.1 (npm), weaviate-agents 1.6.0; Python, Go, Java, C# clients.
28 methods across its docs
llms.txt
YesAgent payments
None foundThe checkout: Reached the card form and stopped: Shared, $45 a month; Checkout: a Stripe card form; An invisible bot check loaded at checkout; Wallets: Google Pay, another wallet.
Docs
No gaps in what we readCovers getting started, auth, error codes and code examples.
9 of 13 documented ways to do the core work worked when tried.
485 of 487 pages read
Tips for agent builders
- After the first console key, create scoped keys with POST /v1/users/db/{id} and assign roles with POST /v1/authz/users/{id}/assign.
- Search with POST /v1/search/{collection}/near-text|bm25|hybrid or gRPC via the SDK; GraphQL may answer 410.
- On free clusters leave vectorIndexType unset (only hfresh) and expect one collection.
- REST hybrid fusionType takes ranked or relativeScore.
- Send X-Weaviate-Cluster-Url with Weaviate Embeddings collections.
- Query Agent MCP: https://api.agents.weaviate.io/mcp with Authorization: Bearer <cluster key> and X-Weaviate-Cluster-Url; use initialize (2025-11-25).
- Engram: use https://api.engram.weaviate.io/v1/...; poll GET /v1/runs/{run_id}; send user_id for user-scoped topics.
How Weaviate compares
In Vector databases
1st of 3 Vector databases products our agent got into on its own.
| Product | Level | Find to leave |
|---|---|---|
| L4 | Find: On its own, Get in: On its own, Use: On its own, Pay: Stopped, Leave: On its own. | |
| L4 | Find: On its own, Get in: On its own, Use: On its own, Pay: Partly, Leave: On its own. | |
| L4 | Find: On its own, Get in: On its own, Use: On its own, Pay: Partly, Leave: On its own. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Partly, Use: Not reached, Pay: Not reached, Leave: Not reached. | |
| L1 | Find: On its own, Get in: Stopped, Use: Not reached, Pay: Partly, Leave: On its own. |
Sorted by level, then work done, then time to the first call. How we rank
For the team behind Weaviate
Our agent got through Pay only in part. Prove Weaviate is yours, then see where each issue is, read your safety findings and ask for a free drive after a fix. How to fix each one comes with the Audit.
8 things to fix for agents
Agent blockers: 2 foundAgent surfaces: 22 issues foundSafety review: 2 issues found
- FindClear
- Get inClear
- UseClear
- PayPartly
- LeaveClear
On our standard
23 criteria: against other vector databases products where we measured 5 or more, otherwise against every product we measured.
- Every way in works for an agentAhead of most 4 of 5 others (vector databases products) lower, 1 the same.
- The best way in is built for agentsAhead of most 4 of 5 others (vector databases products) lower, 1 the same.
- Structured resultsAhead of most 26 of 36 others (every product) lower, 10 the same.
- Results sized for a context windowAhead of most 16 of 31 others (every product) lower, 15 the same.
- Agents pick the right toolBehind most 3 of 4 others (vector databases products) higher, 1 the same.
- Docs an agent can navigateBehind most 4 of 5 others (vector databases products) higher, 1 the same.
- A request id on every answerBehind most 4 of 5 others (vector databases products) higher, 1 the same.
- Fast to the first callThe lowest result All 5 others (vector databases products) higher.
All 23 criteria
- Truthful annotationsIn the middle 2 of 4 others (vector databases products) lower, 2 the same.
- Answers sized for an agentLevel with most 3 of 4 others (vector databases products) the same, 1 lower.
- Prices an agent can readLevel with most 4 of 5 others (vector databases products) the same, 1 lower.
- Errors say what to do nextLevel with most 4 of 5 others (vector databases products) the same, 1 lower.
- Export before leavingIn the middle 2 of 5 others (vector databases products) lower, 2 the same, 1 higher.
- A path written for agentsLevel with most 3 of 5 others (vector databases products) the same, 1 lower, 1 higher.
- Tool names that sort themselvesThe same as all 4 others (vector databases products).
- Typed, tight input schemasIn the middle 2 of 4 others (vector databases products) lower, 2 higher.
- A tool list that fitsThe same as all 4 others (vector databases products).
- Real status codesThe same as all 5 others (vector databases products).
- Spend an agent can see and capThe same as all 5 others (vector databases products).
- WebMCP tools on the web appThe same as all 5 others (vector databases products).
- Pages fit one fetchLevel with most 4 of 5 others (vector databases products) the same, 1 higher.
- Errors a program can readBehind most 3 of 5 others (vector databases products) higher, 2 the same.
- Descriptions that say when to use the toolBehind most 3 of 4 others (vector databases products) higher, 1 the same.
Where Weaviate fits
Found an error? Tell us on this page or read how corrections work.
Questions people ask about Weaviate
- Can an AI agent sign up for Weaviate on its own?
- Yes, in a browser, without a human. console.weaviate.cloud: email, 6-digit code, password. A first address hit the product's limit "Too many sign-in emails have been sent to this address" after our agent's browser tool reopened the page and requested codes repeatedly; a second address went through in one session. No bot check, phone or card.
- Does Weaviate have an MCP server, and can an agent connect to it?
- Yes. Weaviate has an official MCP server. A new agent can get in on its own: in the drive of 28 Oct 2026, our agent signed up, verified and got its key without a person. Through the MCP server, 3 of our agent's 3 attempts worked.
- How does an AI agent get a Weaviate API key?
- In the drive of 28 Oct 2026, our agent got its Weaviate API key in a browser, without a person.
- Can an AI agent delete its Weaviate account?
- Yes, in a browser, without a human. Account page > Delete Account (reason required) signs out; "Your Weaviate account was deleted" mail in 4 minutes; password sign-in gone; Engram key 401 after about 6 minutes; cluster no longer served after about 14 minutes. Browser only, no API.
- Can an AI agent pay for Weaviate?
- Partly. Reaches the checkout: stopped at the $45 a month charge. No card needed to start.
Cite this page
The page as a reference, in plain text, APA and BibTeX.
Cite this
AgentDrives, “Weaviate: AI agent test drive”, 28 October 2026, https://agentdrives.com/tools/weaviate. CC BY 4.0.
AgentDrives. (2026, October 28). Weaviate: AI agent test drive. https://agentdrives.com/tools/weaviate
@misc{agentdrives_weaviate_2026,
author = {{AgentDrives}},
title = {Weaviate: AI agent test drive},
year = {2026},
month = oct,
url = {https://agentdrives.com/tools/weaviate},
note = {Agent test drive, CC BY 4.0},
}Something wrong on this page?
Tell us what's wrong. We check it against the drive and correct the page, with the date.