Skip to main content

​
What HydraDB is

HydraDB is a unified context layer for your AI. One brain for all your working context:
  • Memories: Preferences, decisions, and past conversations that persist across sessions.
  • Knowledge: Documents, app sources, and facts your AI reasons over.
Everything lives in a database you create. You ingest your context and connect the tools your team already uses; HydraDB builds a context graph from it and indexes it in the background. It is serverless: you never provision, shard, or scale anything, and the same code runs at ten documents or ten million. New here? The Quickstart takes you from an empty database to your first query in five minutes.

​
The problem we’re solving

VectorDBs find what’s similar. But your agents want what’s useful.
A useful answer may depend on who made a decision, how a project changed, and who is asking. A sales rep searching for “project Acme” needs the latest sales deck; an engineer needs the changelog and architecture decisions. Similarity alone does not account for those differences. To handle this, teams often combine a vector database, a graph database, a relational store, and a cache. They build the connections between them, keep data in sync, and maintain the retrieval logic themselves. HydraDB keeps your knowledge, user memories, and the relationships between them in one graph, so your agent can retrieve the context that fits the person asking.

​
The principle

HydraDB gives you primitives, not a fixed product. The graph, the memory primitives, the retrieval pipeline, and the ranking knobs are yours to compose into your own context stores, memory layers, and workflows. Your context. Your opinions.

​
Performance

  • Memory recall: 90.79% on LongMemEval-S.
  • Company knowledge: 88.7 overall on EnterpriseRAG-Bench, retrieval across more than half a million documents from nine enterprise systems.
  • Coding harness: 90.73 on CodeWikiBench for nlohmann/json, against 66.06 for DeepWiki and 61.28 for CodeWiki.
  • Long-term memory at scale: 82% overall on BEAM 1M, across ten dimensions.
  • Financial AI: 91.4% recall@10 on FinanceBench, over dense filings and enterprise data.
  • Low latency: sub-200 ms retrieval.
The full technical breakdown of each run is in our benchmarks.

​
More in the box

  • Connectors: 70 providers, including Slack, Gmail, Google Drive, Notion, GitHub, Jira, Confluence, Linear, HubSpot, Zendesk, and Microsoft Teams, synced on a schedule. Where the provider supports it, the source app’s permissions come along as access control, so a query that sends a user’s identity as acl only returns what that user may see. See Connectors and Access Control.
  • SDKs: Python and TypeScript, with a helper that turns a query result into a prompt string.
  • Tools: an MCP server for Claude, Cursor, and other MCP clients; a CLI; and memory plugins for coding agents and agent frameworks such as Claude Code, Codex, Hermes, and OpenClaw.

​
What you can build

  • Customer support agents grounded in real customer history and preferences
  • Coding agents with persistent memory of a codebase and team conventions
  • Clinical companions tracking patient context across visits
  • Research copilots reasoning across papers, authors, and findings
  • Internal knowledge assistants spanning Slack, Notion, Drive, and email
  • Consumer AI apps where every user gets a second brain that evolves over time
That is a sample, not the limit. The Cookbooks walk through a dozen of these end to end.

​
Get started

  1. Sign up at dashboard.hydradb.com for your API key.
  2. Follow the Quickstart.
  3. Explore Core Concepts and the API Reference.
For enterprise onboarding, contact [email protected].

​
For AI agents

For AI coding agents and IDE assistants, use the HydraDB Agent Integration Guide and the v2 OpenAPI spec.