turbopuffer
turbopuffer is a serverless search engine that combines vector and full-text (BM25) search built from first principles directly on object storage. It exposes a single REST API organized around namespaces — each namespace stores documents with vector embeddings, attributes, and full-text indexes — and supports approximate nearest neighbor, full-text BM25, and hybrid query patterns with attribute filtering, ranking, and aggregation. The platform is used in production by Anthropic, Cursor, Notion, Linear, Superhuman, Pylon, Readwise, and Telus, and handles 4T+ documents, 10M+ writes/s, and 25k+ queries/s across customer fleets. Official client libraries ship for Python, TypeScript, Go, Java, Ruby, and C#, generated from a public OpenAPI 3.1 specification via Stainless. Pricing is tiered (Launch, Scale, Enterprise) with usage-based metering on storage, writes, and queries on top of monthly minimums.
11 APIs
0 Features
Vector SearchFull-Text SearchHybrid SearchBM25ServerlessObject StorageRAGSemantic SearchAI InfrastructureEmbeddings
Single REST API for all turbopuffer operations — namespace metadata, write (upsert / patch / delete), vector ANN query, full-text BM25 query, hybrid query, multi-query, explain_...
Endpoints for upserting, patching, and deleting documents within a namespace. Writes are batched into a per-namespace write-ahead log and become queryable once committed to obje...
Unified query endpoint that runs vector ANN, full-text BM25, and hybrid queries against a namespace, with attribute filters, top-k, aggregation groups, and ranking controls. Sup...
Namespace lifecycle and metadata endpoints — list namespaces, read schema / dimensions / row count, warm the cache, export contents, branch_from (copy-on-write clones in constan...
Official Python client library for the turbopuffer REST API, Stainless-generated from the public OpenAPI spec.
Official TypeScript / JavaScript client library for the turbopuffer REST API, Stainless-generated and published to npm.
Official Go client library for the turbopuffer REST API, Stainless-generated from the public OpenAPI spec.
Official Java / Kotlin client library for the turbopuffer REST API, Stainless-generated from the public OpenAPI spec.
Official Ruby client library for the turbopuffer REST API, Stainless-generated from the public OpenAPI spec.
Official C# / .NET client library for the turbopuffer REST API, Stainless-generated from the public OpenAPI spec.
Open-source general-purpose benchmarking tool for turbopuffer deployments. Useful for validating recall, latency, and throughput on a given workload.
aid: turbopuffer
url: https://raw.githubusercontent.com/api-evangelist/turbopuffer/refs/heads/main/apis.yml
name: turbopuffer
kind: company
description: >-
turbopuffer is a serverless search engine that combines vector and full-text
(BM25) search built from first principles directly on object storage. It
exposes a single REST API organized around namespaces — each namespace stores
documents with vector embeddings, attributes, and full-text indexes — and
supports approximate nearest neighbor, full-text BM25, and hybrid query
patterns with attribute filtering, ranking, and aggregation. The platform is
used in production by Anthropic, Cursor, Notion, Linear, Superhuman, Pylon,
Readwise, and Telus, and handles 4T+ documents, 10M+ writes/s, and 25k+
queries/s across customer fleets. Official client libraries ship for Python,
TypeScript, Go, Java, Ruby, and C#, generated from a public OpenAPI 3.1
specification via Stainless. Pricing is tiered (Launch, Scale, Enterprise)
with usage-based metering on storage, writes, and queries on top of monthly
minimums.
image: https://kinlane-images.s3.amazonaws.com/shared/apis-json/apis-json-logo.jpg
tags:
- Vector Search
- Full-Text Search
- Hybrid Search
- BM25
- Serverless
- Object Storage
- RAG
- Semantic Search
- AI Infrastructure
- Embeddings
created: '2026-05-23'
modified: '2026-05-25'
specificationVersion: '0.19'
apis:
- aid: turbopuffer:rest-api
name: turbopuffer REST API
description: >-
Single REST API for all turbopuffer operations — namespace metadata,
write (upsert / patch / delete), vector ANN query, full-text BM25 query,
hybrid query, multi-query, explain_query, branch_from, copy_from, and
cache warming — with bearer-token authentication. The base host is
region-templated as https://{region}.turbopuffer.com (e.g.,
gcp-us-east4.turbopuffer.com or aws-us-east-1.turbopuffer.com) to
minimize egress and latency.
humanURL: https://turbopuffer.com/docs
baseURL: https://api.turbopuffer.com
tags:
- REST
- Vector Search
- Full-Text Search
- Hybrid Search
properties:
- type: Documentation
url: https://turbopuffer.com/docs
- type: Authentication
url: https://turbopuffer.com/docs/auth
- type: Quickstart
url: https://turbopuffer.com/docs/quickstart
- type: OpenAPI
url: https://raw.githubusercontent.com/api-evangelist/turbopuffer/main/openapi/turbopuffer-openapi.yml
- type: OpenAPI-Source
url: https://github.com/turbopuffer/turbopuffer-openapi
- aid: turbopuffer:write
name: turbopuffer Write API
description: >-
Endpoints for upserting, patching, and deleting documents within a
namespace. Writes are batched into a per-namespace write-ahead log and
become queryable once committed to object storage. Supports both
row-oriented and column-oriented batch formats.
humanURL: https://turbopuffer.com/docs/write
baseURL: https://api.turbopuffer.com
tags:
- Write
- Upsert
- Namespaces
properties:
- type: Documentation
url: https://turbopuffer.com/docs/write
- aid: turbopuffer:query
name: turbopuffer Query API
description: >-
Unified query endpoint that runs vector ANN, full-text BM25, and hybrid
queries against a namespace, with attribute filters, top-k, aggregation
groups, and ranking controls. Supports multi-query (up to 16 per
request) and explain_query for query planning.
humanURL: https://turbopuffer.com/docs/query
baseURL: https://api.turbopuffer.com
tags:
- Query
- Vector
- BM25
- Hybrid
- ANN
properties:
- type: Documentation
url: https://turbopuffer.com/docs/query
- aid: turbopuffer:namespaces
name: turbopuffer Namespaces API
description: >-
Namespace lifecycle and metadata endpoints — list namespaces, read
schema / dimensions / row count, warm the cache, export contents,
branch_from (copy-on-write clones in constant time), copy_from, and
delete a namespace.
humanURL: https://turbopuffer.com/docs/namespaces
baseURL: https://api.turbopuffer.com
tags:
- Namespaces
- Metadata
- Admin
- Branching
properties:
- type: Documentation
url: https://turbopuffer.com/docs/namespaces
- aid: turbopuffer:python-sdk
name: turbopuffer Python SDK
description: >-
Official Python client library for the turbopuffer REST API,
Stainless-generated from the public OpenAPI spec.
humanURL: https://github.com/turbopuffer/turbopuffer-python
baseURL: https://github.com/turbopuffer/turbopuffer-python
tags:
- SDK
- Python
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-python
- type: Package
url: https://pypi.org/project/turbopuffer/
- aid: turbopuffer:typescript-sdk
name: turbopuffer TypeScript SDK
description: >-
Official TypeScript / JavaScript client library for the turbopuffer REST
API, Stainless-generated and published to npm.
humanURL: https://github.com/turbopuffer/turbopuffer-typescript
baseURL: https://github.com/turbopuffer/turbopuffer-typescript
tags:
- SDK
- TypeScript
- JavaScript
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-typescript
- type: Package
url: https://www.npmjs.com/package/@turbopuffer/turbopuffer
- aid: turbopuffer:go-sdk
name: turbopuffer Go SDK
description: >-
Official Go client library for the turbopuffer REST API,
Stainless-generated from the public OpenAPI spec.
humanURL: https://github.com/turbopuffer/turbopuffer-go
baseURL: https://github.com/turbopuffer/turbopuffer-go
tags:
- SDK
- Go
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-go
- aid: turbopuffer:java-sdk
name: turbopuffer Java SDK
description: >-
Official Java / Kotlin client library for the turbopuffer REST API,
Stainless-generated from the public OpenAPI spec.
humanURL: https://github.com/turbopuffer/turbopuffer-java
baseURL: https://github.com/turbopuffer/turbopuffer-java
tags:
- SDK
- Java
- Kotlin
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-java
- aid: turbopuffer:ruby-sdk
name: turbopuffer Ruby SDK
description: >-
Official Ruby client library for the turbopuffer REST API,
Stainless-generated from the public OpenAPI spec.
humanURL: https://github.com/turbopuffer/turbopuffer-ruby
baseURL: https://github.com/turbopuffer/turbopuffer-ruby
tags:
- SDK
- Ruby
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-ruby
- aid: turbopuffer:csharp-sdk
name: turbopuffer C# SDK
description: >-
Official C# / .NET client library for the turbopuffer REST API,
Stainless-generated from the public OpenAPI spec.
humanURL: https://github.com/turbopuffer/turbopuffer-csharp
baseURL: https://github.com/turbopuffer/turbopuffer-csharp
tags:
- SDK
- CSharp
- DotNet
properties:
- type: Repository
url: https://github.com/turbopuffer/turbopuffer-csharp
- aid: turbopuffer:benchmark
name: tpuf-benchmark
description: >-
Open-source general-purpose benchmarking tool for turbopuffer
deployments. Useful for validating recall, latency, and throughput on a
given workload.
humanURL: https://github.com/turbopuffer/tpuf-benchmark
baseURL: https://github.com/turbopuffer/tpuf-benchmark
tags:
- Tooling
- Benchmark
- Go
properties:
- type: Repository
url: https://github.com/turbopuffer/tpuf-benchmark
common:
- type: Website
url: https://turbopuffer.com/
- type: Documentation
url: https://turbopuffer.com/docs
- type: GitHub
url: https://github.com/turbopuffer
- type: OpenAPI
url: https://github.com/turbopuffer/turbopuffer-openapi
- type: Pricing
url: https://turbopuffer.com/pricing
- type: Architecture
url: https://turbopuffer.com/docs/architecture
- type: Regions
url: https://turbopuffer.com/docs/regions
- type: Limits
url: https://turbopuffer.com/docs/limits
- type: Blog
url: https://turbopuffer.com/blog
- type: LLMsTxt
url: https://turbopuffer.com/llms.txt
- type: TermsOfService
url: https://turbopuffer.com/terms-of-service
- type: Customers
url: https://turbopuffer.com/customers
integrations:
- name: Anthropic
description: >-
Anthropic is a publicly listed turbopuffer customer, using the platform
as serverless vector + full-text search infrastructure for AI workloads
alongside Anthropic's Claude model family.
url: https://turbopuffer.com/
- name: Cursor
description: >-
Cursor uses turbopuffer at >1T vectors and 80M+ namespaces to power
semantic code retrieval inside the Cursor AI coding environment,
reporting up to 23.5% improvement in agent accuracy and a 20x cost
reduction versus prior infrastructure.
url: https://turbopuffer.com/customers/cursor
- name: Notion
description: >-
Notion uses turbopuffer as the search backbone for workspace content
across its customer base.
url: https://turbopuffer.com/customers/notion
- name: Linear
description: >-
Linear uses turbopuffer to power search across issues, projects, and
docs at scale.
url: https://turbopuffer.com/customers/linear
- name: Superhuman
description: >-
Superhuman uses turbopuffer as part of its AI-assisted email search and
retrieval stack.
url: https://turbopuffer.com/customers
- name: Stainless
description: >-
Stainless generates turbopuffer's official client libraries (Python,
TypeScript, Go, Java, Ruby, C#) directly from the public OpenAPI 3.1
specification.
url: https://www.stainless.com/
maintainers:
- FN: Kin Lane
email: kin@apievangelist.com