For your product
A familiar path to ship grounded answers: index your knowledge, drop in the SDK, and keep the UI focused on the conversation.
Grid
Retrieval that keeps up with conversation
An in-process search runtime for voice agents and copilots—fast enough that context never stalls the turn.
Trusted by teams building real-time AI
- Northline
- Cascade Labs
- Harbor AI
- Pulsekit
- Vertex Soft
Your agent’s source of context
Humans and agents share one retrieval layer—loaded beside the runtime, ready before the next sentence starts.
For your agents
Agents query locally with full project context—no waiting on a remote hop every time they need a fact.
One index. Every surface your agent touches.
Package knowledge once. Grid distributes and refreshes it wherever the session runs—browser, edge, or server.
- Browser & WASM
- Node & Python
- On-device sessions
- Cloud sync
Ask less. Retrieve more.
Grid sits under the chat—not another console of filters. Your agent embeds, searches, and returns ranked context as a local call.
What changed in the refund policy last week?
3 matches · 6 ms · hybrid semantic + keyword
Policy v4.2 — Restocking waived under 14 days…
What you can do
Run the day
Ground FAQs, docs, and product facts mid-call without pausing for a network round trip.
Grow the product
Add session indexes on the fly—short-lived context that syncs when the conversation ends.
Work everywhere
Same API in the browser, on a phone, or behind your agent server. Swap surfaces without rewriting retrieval.
Turn knowledge into something agents can use
Point Grid at your corpus. It builds a compact index, ships it to the runtime, and keeps it fresh with hot swaps—no cluster to babysit.
- 01 Connect sources and describe what matters
- 02 Grid embeds and packs a portable index
- 03 Agents load it in-process and query in milliseconds
A simple interface on a sharper engine
Indexes, sessions, and filters—structured enough for teams, fast enough for live dialogue.
Everything you need to ground agents
Built for teams shipping conversational products.
-
Hybrid retrieval
Semantic and keyword search in one pass, with metadata filters.
-
Local embeddings
Bundle compact models or bring your own vectors.
-
Live sessions
Index during a call, then sync when you’re ready.
-
Portable runtime
Rust core with WASM—browser to backend without a rewrite.
-
Developer API
Wire Grid into the agent framework you already use.
-
Quiet ops
No shards to tune. Load an index, query, move on.
Ship agents that feel present
Skip the retrieval lag. Start with Grid and keep every turn in the conversation.