Combinator
Founders, Inc.AGENT MEMORY
Pick up where you left off.
Bring the right history into the next conversation.
Explore memoryLoading page
Intelligence is only as good as its context.
Graph, vector, and text retrieval over your Postgres.
We virtualise the GraphRAG layer. You keep shipping.
THE PATH OF LEAST COMPLEXITY
Graph, vector, and text retrieval. One path from your database to your AI.
MORE SIGNAL. EVERY MILLISECOND.
Your AI has a time budget.
Give it more of the right context.
2.44 ms measured latency
2.56 ms measured latency
21% more recall within 3 ms. 91% vs 75%, a 16-point lift.
Explore the benchmarkMeasured results · GloVe 1.18M · Recall@10 · p50 latency · Best observed points within budget
RETRIEVAL, IN EVERY DIMENSION
Find the meaning behind the query. Vector search connects the words your users choose with the information they need.
Explore the SDKpage = project.vector.search(
query_embedding,
config="documents_embedding",
limit=10,
)Use an embedding from your configured model.
YOUR DATABASE. MORE DIMENSIONS.
Search by meaning, follow relationships, and retrieve the context your AI needs. All on the database you already know.
Explore the platformGraphTraverse connected records, one relationship at a time.
BUILT AROUND YOUR POSTGRES
Connect your existing Postgres to Polygres, or start with a database hosted by us.
polygresHost on Polygres
Connect existing
Connect existing
PlanetScaleConnect Postgres
PostgreSQLBring your databaseMADE FOR THE PEOPLE BUILDING IT
Explore in the playground. Automate with the CLI. Build retrieval into your application with the Python SDK.
Open the documentationWHAT WILL YOU BUILD?
AGENT MEMORY
Bring the right history into the next conversation.
Explore memoryGROUNDED ANSWERS
Retrieve the evidence. Keep the connections.
Try retrievalPRODUCT SEARCH
Turn everyday language into relevant discoveries.
Explore semantic searchBUILT IN THE OPEN
pgGraph and pgContext are the open-source engines behind Polygres. Inspect how they work. Run them yourself. Build on infrastructure you can understand.
Meet the people behind itOUR THESIS
Itâs a virtual layer between your data and the intelligence that needs it. Context that takes shape around a question, then makes way for the next. Weâre building toward that future by solving AI retrieval.
Read our manifestoPolygres is an AI retrieval platform built on Postgres. It brings graph traversal, vector similarity, full-text search, and hybrid retrieval into a platform developers can explore in a dashboard and use through APIs and the Python SDK.
Semantic search uses vector similarity to find related meaning. Full-text search matches language through a text index. Both are useful: the right approach depends on whether the question needs meaning, specific words, or both.
It is our thesis for AI-native data: useful context should take shape around the question. The underlying data remains durable. The context an application retrieves can change from one request to the next. We are building toward that future through retrieval on Postgres.
pgGraph and pgContext, the Postgres extensions behind Polygres, are open source. You can inspect their code, read their benchmarks, and run the extensions yourself. The managed platform is available through Polygres.
BUILD THE NEXT GENERATION OF AI.