Skip to main content

What is Ground?

Ground is a retrieval and indexing platform for coding agents: it ingests repos, documentation, OpenAPI specs, PDFs, and public packages, then serves hybrid search (vectors + keywords), RMH search (multi-round / neighbor expansion), streaming chat with citations, Rover (GitHub exploration jobs), and on-demand package search.
  • Provenance: Results carry citations (source, path, lines, version where applicable).
  • Freshness: Sources have sync metadata; policy controls how stale content is treated.
  • Quality signals: Evidence counts, staleness summaries, and optional conflict surfacing for divergent definitions.

Why Ground?

  1. Index your sources (Git, HTML docs, OpenAPI, npm/PyPI, PDFs).
  2. Query with hybrid + RMH when you need more than a single top-k list.
  3. Wire agents through the REST API or MCP using your API key.

Quick Start

Create a source and run your first search in minutes

Concepts

Sources, jobs, search, citations, freshness

MCP

Connect Ground to MCP-compatible clients

API Reference

Full REST surface (search, chat, Rover, packages, keys)

Key features

  • Git: language-aware chunking
  • Docs: sitemap or URL lists
  • OpenAPI: operations and schemas
  • Packages: npm / PyPI (catalog + on-demand search)
  • PDF: upload pipeline with page-aware chunks
REST, MCP (https://api.trygroundai.com/mcp), and official TypeScript / Python SDKs.
Configure staleness budgets, source weights, and minimum evidence thresholds via /policy.

Documentation feedback

Email ali@trygroundai.com.