Grill is POMA AI's Managed Context Engine — an end-to-end RAG pipeline you drop into your agent or application.
It's built for teams shipping agents, copilots, or document-search products who want best-in-class retrieval as a fully-managed service — we handle the devops and MLops. Plug into your agent through MCP, SDK, or REST — retrieval from day one.
Grill is a full, end-to-end RAG pipeline. Simply sign-up and you can start ingesting and querying documents immediately from your AI agent or pipeline, all using our best-in-class structural chunking and query system.
Getting started is easy. Simply sign up and, if you want, set up an organisation and project to share with others. You can start querying your documents in minutes.
Documents can be ingested manually through the console, by URL, via our API, our SDK, our CLI, or directly in your AI agent using an MCP server.
As documents are ingested, they are chunked and organised using our patented structurally-aware chunking technique.
As soon as documents are chunked they become available for querying through our SDK, API, or MCP servers immediately.
Add the Grill MCP server to your agent's configuration, drop in your project key, and you're ready to query. No vector DB to provision, no embeddings to manage - just a context endpoint your agent can call.
Your agent's two core tools are ingest and query. Ingest accepts 50+ filetypes - PDFs, DOCX, HTML, images, spreadsheets, and more. Query searches across text, visual content, and tables, diagrams, embedded content, and more, and in any of our supported languages.
{
"mcpServers": {
"poma-grill": {
"command": "npx",
"args": [
"-y",
"@poma-ai/poma-grill-mcp",
"-input",
"-"
],
"env": {
"POMA_API_KEY": "sk_..."
}
}
}
}To get started all you need is to sign up, grab your Grill project key, and start ingesting your documents through your AI agent.