Home / Engine / Alternatives / turbopuffer

StackGrep Engine vs turbopuffer

turbopuffer is serverless vector and full-text search built on object storage, aimed at large numbers of namespaces. StackGrep Engine is something else: exact text and regex search over your own documents, built for AI agents, with every match counted and your index kept in object storage.

Join the waitlist See the engine

What StackGrep does differently

Regex over whole documents, not just ranked words and vectors
No embeddings needed: push text and search it
Exact counts of every match
A hosted MCP server and S3 bucket sync

When turbopuffer is the better pick

If you need vector search or ranked hybrid search at scale, and you're happy to compute embeddings.

Three calls to try it

POST /api/collections/docs/docs     {"docs": [{"id": "a1", "text": "...", "meta": {"team": "ops"}}]}
GET  /api/collections/docs/search?q=net+30&filter=team:ops
GET  /api/collections/docs/count?q=net+30

Quickstart →

Questions

Is StackGrep built the same way?

Both keep data in object storage and cache it where it's searched, so idle data is cheap. StackGrep indexes text for exact and regex matching; it has no vectors.

Do I need embeddings?

No. Push text and search it. If you also want semantic search, keep vectors elsewhere and use StackGrep for exact lookups.

Other comparisons

Elasticsearch alternative Algolia alternative Typesense alternative Meilisearch alternative Pinecone alternative

Use cases

Hosted grep for your AI agents Exact counts for RAG and agents Exact match search for AI agents RAG without embeddings Grep an S3 bucket: search the contents of every file A regex search API that doesn't scan everything A knowledge base MCP server for your documents Find every contract that says exactly this

Give your agents exact search

StackGrep Engine is onboarding teams from the waitlist. Bring your documents or your bucket.

Join the waitlist Read the docs