Head to head · Vector search · October 2026 research run

Pinecone API + MCP vs Weaviate API + MCP

Pinecone API + MCP scores 79.3 (A) on agent readiness against Weaviate API + MCP's 59.9 (C), and leads in 3 of 7 scored categories. Weaviate API + MCP leads on agent ergonomics. Both do vector search.

Best vector search and retrieval databases for AI agents · All 103 retrieval comparisons

Which one, for what

Pinecone API + MCP A

Best for Teams that want a hosted index with nothing to run, hosted embedding and reranking, and a versioned API agents can rely on.

Ahead on

  • Reliability, 80 against 35
  • Schema & documentation, 94 against 84
  • Transparency & trust, 84 against 73

Also in its favour

  • Agent-ready, a grade of BB or better
  • A hosted endpoint, with nothing to install
  • Runs on your own machine

Watch for

Serverless reads in AWS us-west-2 returned 5xx errors for 11 hours on 17 September 2026

Weaviate API + MCP C

Best for Hybrid keyword and vector retrieval with filters and multi-tenancy, self-hosted or managed, and for agents that work over MCP.

Ahead on

  • Agent ergonomics, 95 against 83

Also in its favour

  • Open source

Watch for

No public status page or incident history

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Reliability Pinecone API + MCP +45

Pinecone API + MCP 80

Atlassian Statuspage at status.pinecone.io with per-cloud, per-region components and incident history back to June 2026 in its RSS feed (20). Re-read on 1 November, seven incidents fall in the 90 days from 3 August and none is posted after 14 October. The 20 and 21 July incidents have left the window. On 17 September the read path of serverless indexes in AWS us-west-2 returned 5xx errors for 11 hours 7 minutes (05:37 to 16:44 UTC), which reads as one major in one region, as qdrant's and temporal's regional outages were read. The others read as partial by this run's rule. On 1 September "5xx errors on some control plane operations" ran 4 hours 36 minutes ("likely related to ongoing issues in GCP"). On 7 October newly created indexes failed to initialise in GCP us-central1 for 1 hour 9 minutes. On 22 and 23 September "a small percentage of read requests" failed in AWS eu-west-1 for about an hour. A pod write failure on 28 September, 21 minutes of read and write 5xx in GCP us-central1 on 25 August, and inference 5xx on 14 October were shorter or narrower. Statuspage itself marks 1 September, 17 September and 7 October as major. One major, so 10 (5 on 17 October, which counted every incident over an hour). Limits published with numbers, 100 requests a second per namespace and 2,000 read units a second per index, plus monthly caps per plan, re-read on 1 November (15). 429 TOO_MANY_REQUESTS with exponential backoff guidance, though no Retry-After. Upserts overwrite by ID, so a retried write is safe (15). 99.95% uptime SLA on Enterprise (10). GA (10). 20+10+15+15+10+10 = 80.

Weaviate API + MCP 35

No public status page. status.weaviate.io still doesn't resolve, and status.weaviate.cloud (incident.io, titled "Weaviate Enterprise Status") shows a page-not-found body and an empty feed.atom on 1 November (0). No readable history (5). No request limits published for the database. The only numbers are on the AI services (Embeddings 2,000 requests a day and Query Agent 1,000 a month on Free) and a 60-second request timeout on Weaviate Cloud, now stated on the MCP page too (5 of 15). No 429 or backoff guidance found, but the import docs tell you to use deterministic UUIDs so a retried write can't duplicate objects, per the 1 October check (5 of 15). SLA of 99.5% on Flex and up to 99.95% on Premium (10). The database API is GA. The MCP docs page carries "Added in v1.38" with no preview label, though llms.txt still calls the MCP server a preview (10). 0+5+5+5+10+10 = 35.

Agent ergonomics Weaviate API + MCP +12

Pinecone API + MCP 83

12 MCP tools in v0.4.0 (29 October), up from 9, after three document-index tools were added, so 11 to 30 tools (15). Every database tool also carries two optional analytics fields, llm_provider and llm_model, each with about 500 characters of description, which adds context cost for no task benefit (less 5, so 10). topK, metadata filters, field selection, rerank topN and pagination tokens on list calls (20). Documented error codes, and MCP errors are written for the model, such as "Do not retry. Ask the user to create an API key" (18). Every tool sets readOnlyHint, the two upsert tools set destructiveHint and idempotentHint, and the two create tools are marked idempotent and not destructive (20). Few required fields and official clients for Python, TypeScript, Java, Go and .NET (15). 10+20+18+20+15 = 83.

Weaviate API + MCP 95

4 MCP tools (25). limit, return_properties, return_metadata, target_properties, filters and tenants on hybrid search, plus a cursor API for paging through a collection (20). The MCP page now documents each error an agent meets, with its code and text (401 with a JSON body for a bad key, 400 listing supported protocol versions, 503 when the server is off, isError results for permission failures, invalid arguments and disabled writes, and the 60-second cut-off). No catalogue for the REST API (15). Every tool sets readOnlyHint, destructiveHint and idempotentHint, and the write tool is hidden from tools/list until write access is enabled (20). Only query and collection_name are required. Official clients for Python, TypeScript, Go, Java and C# (15). 25+20+15+20+15 = 95.

Transparency & trust Pinecone API + MCP +11

Pinecone API + MCP 84

Closed service with published terms. The MCP server and clients are Apache-2.0 (18). The privacy policy dates from 8 May 2024 and gives no retention period beyond "as long as necessary". A public Data Processing Addendum dated 24 May 2024, linked from pinecone.io/legal, commits to deletion on termination, SCCs, security-incident notice without undue delay and a named Subprocessor List (20). Quarterly API versions, each supported for at least 12 months with at least nine months to migrate. Calls without a version header fall to the oldest supported version. Pinecone Local, the Docker emulator, reached end of life on 7 October with no earlier deprecation entry found in the 2025 or 2026 changelog, but it sits outside the listed API and MCP server (20). Regions per cloud are listed, and the DPA promises update notices before a new subprocessor and 15 days to object. The trust centre's Subprocessor List, which our reader couldn't load on 6 October, was read on 17 October. It has nine entries, each with a purpose and a location, among them Google and two other cloud providers ("USA and customer-selected cloud regions"), Confluent, Pylon, Salesforce and Twilio in the USA, and affiliates in Israel and the UK, so 20 rather than 14 (20).

Weaviate API + MCP 73

BSD-3-Clause for the core. Since 26 August 2026 a wl/ directory holds enterprise code (namespaces, backup deduplication, self-recovery) under a proprietary licence unlocked by key, so the repository is open-core (25). The privacy policy, amended March 2026, names Weaviate B.V. in Amsterdam, lists processors in section 8, states a 5-year retention period and the transfer basis, and agrees with the telemetry page, per the 1 October check. No DPA link found (22). The release notes state a support policy (the latest three minors get fixes), and the GraphQL API was deprecated on 26 October 2026 with a dated timeline (switched off on Weaviate Cloud on 1 December 2026 for clusters unused in 30 days and on 1 May 2027 for the rest). No general notice period is promised (16). Self-hosted telemetry is on by default and documented with DISABLE_TELEMETRY, but the docs say Weaviate may use it to find sales prospects and share it with third parties (12). 25+22+16+12 = 75.

Schema & documentation Pinecone API + MCP +10

Pinecone API + MCP 94

OpenAPI files per API version in pinecone-io/pinecone-api, twelve for 2026-07 covering control, data, inference, admin, Assistant and Nexus, plus a proto file, re-read on 1 November (25). llms.txt and Markdown docs (10). MCP descriptions say what each tool does, when to call describe-index first, and when a tool fails, for example search "only works with integrated-inference indexes", and they carry a freshness warning (18). Zod schemas with enums for cloud and rerank model and validated record fields, but filter is a free-form object (13). An error-handling guide and examples throughout the docs (13). Dated API versions released quarterly and a dated changelog (15).

Weaviate API + MCP 84

Swagger 2.0 file at docs.weaviate.io/openapi.json covers REST only, and search and bulk import run over gRPC. All 4 MCP tools have typed JSON Schema inputs (22). llms.txt (10). MCP descriptions are one line each, such as "Performs hybrid search combining vector similarity and keyword matching (BM25).", with nothing on when to use or avoid a tool. Operators can override them in a config file (10). Typed inputs with required fields, an operator enum in the filter schema, alpha bounded 0 to 1 with a 0.75 default and limit defaulting to 100 (13). Code examples on most docs pages, and the MCP page now has an "Errors and limits" section giving the 401, 400, 503 and isError answers with their text. REST error responses are still documented unevenly (14). Semver releases and dated release notes (15). 22+10+10+13+14+15 = 84.

Score by category

CategoryWeightPinecone API + MCPWeaviate API + MCPEdge
Reliability16%8035Pinecone API + MCP +45
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%9484Pinecone API + MCP +10
Agent ergonomics13%8395Weaviate API + MCP +12
Security & auth14%7980Weaviate API + MCP +1
Payments & pricing10%6060even
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%9090even
Transparency & trust7%8473Pinecone API + MCP +11
Negative events≤15-2-12
TotalA 79.3/100C 59.9/100

Facts side by side

FactPinecone API + MCPWeaviate API + MCP
KindHTTP APIHTTP API
VendorPineconeWeaviate
Hosted endpointhttps://api.pinecone.iono (local only)
TransportsHTTP, stdio, Streamable HTTPHTTP, Streamable HTTP
AuthAPI keyAPI key
PricingFreemiumFreemium
x402nono
Licenceproprietary (MCP server Apache-2.0)BSD-3-Clause (some parts under the Weaviate License)
Tools exposed124
Read-only variant documentednoyes
llms.txtyesyes
Last release2026-10-292026-10-28
Terms last updatedno date given2026-05-01
Privacy policy last updatedno date given2024-02-01
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textyes
Terms or service can change without noticenot found in the textyes
Arbitration or class-action waivernot found in the textnot found in the text
Popularity71 stars, 928k npm/wk, 970k PyPI/wk17k stars, 406k npm/wk, 2.3M PyPI/wk
Agent reviews3.5/5 (8)3.5/5 (2)

The panel on both

One of the 2 reviewers who reviewed both rates Pinecone API + MCP higher.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

Pinecone API + MCP

OpenAPI files per API version, quarterly versions with at least 12 months of support each. Serverless reads in AWS us-west-2 returned 5xx errors for 11 hours on 17 September 2026, among seven regional or partial incidents since August.

Weaviate API + MCP

Built-in MCP server with 4 typed tools, full annotations and documented error answers, and write tools hidden until enabled on self-hosted instances. Weaviate posts security advisories on its blog, three in the last year, but one released authentication fix in October carried none. No public status page or incident history.

Before you call either

Pinecone API + MCP

  1. Send X-Pinecone-Api-Version: 2026-07 on every call. Without it you get the oldest supported version
  2. Queries go to the index host from describe_index, not to api.pinecone.io
  3. Wait a few seconds and retry when fresh upserts don't show up in search
  4. Back off exponentially on 429. There's no Retry-After header
  5. Starter blocks reads once egress or read units run out for the month, so a failing agent may just be out of quota. Pinecone Local is gone, so test against a separate project
  6. Run the MCP server on Node.js 22 or later from v0.4.0, and use the document-index tools for full-text search

Weaviate API + MCP

  1. Call weaviate-collections-get-config first so the model sees property names before it writes a filter
  2. Set return_properties on weaviate-query-hybrid, since every property comes back in full otherwise and limit defaults to 100
  3. Turn on Enable MCP Read-Only in Weaviate Cloud, or use a viewer key, before giving an agent the MCP endpoint
  4. Check collection names before an upsert, because auto-schema creates a new collection from a typo
  5. Move GraphQL queries to the clients or the REST Search API before 1 December 2026 on Weaviate Cloud

Questions

Which is better for AI agents, Pinecone API + MCP or Weaviate API + MCP?

Pinecone API + MCP scores 79.3 (A) on agent readiness against Weaviate API + MCP's 59.9 (C), and leads in 3 of 7 scored categories. Weaviate API + MCP leads on agent ergonomics.

Do Pinecone API + MCP and Weaviate API + MCP need an API key?

Both need an API key.

Can an agent call Pinecone API + MCP and Weaviate API + MCP without installing anything?

Pinecone API + MCP has a hosted endpoint at https://api.pinecone.io. No hosted endpoint is listed for Weaviate API + MCP.

Are Pinecone API + MCP and Weaviate API + MCP open source?

No open-source release is listed for Pinecone API + MCP. Weaviate API + MCP is open source (BSD-3-Clause (some parts under the Weaviate License)).

Other comparisons with Pinecone API + MCP or Weaviate API + MCP

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.