Open Glean

What does Open Glean do?

Open Glean's README calls it "the AI workspace over Hydra DB". You ask a question, the app retrieves context from your Hydra database, writes the answer, and cites its sources. The README lists one composer per question with inline citations and a sources panel, a Deep Research mode that plans a DAG of sub-questions and runs each level in parallel, scope switching across databases and collections, a Context area for memories, files and saved webpages, collections as sub-tenants, a Mindmap of the knowledge graph Hydra builds, a connectors area for syncing third-party knowledge, and bring-your-own-model support for any OpenAI-compatible endpoint.

Note: the open-source code is the Next.js server that proxies requests to the official @hydradb/sdk. The retrieval and the connector framework are in Hydra DB itself, which the README links at hydradb.com. Open Glean the app is Apache-2.0; the Hydra DB service it talks to is a separate product.

Key facts

• Licence: Apache-2.0 on the code. Connector logos in NOTICE are third-party trademarks and are not covered by the Apache grant.
• Stack: Next.js 16, React 19, Tailwind CSS v4, and @hydradb/sdk. The README says no other runtime dependency is used for data or AI.
• Required: a Hydra DB key, created at app.hydradb.com, and an OPEN_GLEAN_SESSION_SECRET of 16 or more characters in production.
• Answer models: any OpenAI-compatible endpoint; a built-in OpenRouter model picker with favourites is included.
• Deep Research: concurrent runs per instance are capped by OPEN_GLEAN_MAX_CONCURRENT_RESEARCH, default 3. The README says each run costs many LLM and retrieval calls.
• Persistence: chats persist to MongoDB when reachable, with a localStorage fallback.
• Security: Hydra and OpenRouter keys are stored in an AES-256-GCM-encrypted, httpOnly session cookie and used server-side. The README says nothing key-shaped is inlined into the JS bundle, localStorage or any client state.
• A stored key is pinned to the endpoint it was stored with; the LLM base URL can only be chosen by a request that supplies its own key.
• Docker: the image uses Next.js standalone output and runs as a non-root user.
• Amazon DocumentDB does not support IAM database authentication; the IAM token path in the code will not work with DocumentDB.

What does it replace, and where does it fall short?

Open Glean replaces Glean for someone who wants a chat over their own documents and connected apps, with cited sources and a deep-research path, and is willing to host the UI themselves. See the open-source alternatives to Glean page.

Where it falls short: the retrieval, the vector store and the connector framework all sit in Hydra DB, which is a paid hosted service. The repository is the UI and the server proxy, not a complete self-hosted stack like Khoj or Onyx. If you want everything on your own hardware with no third-party service, a different pick on the alternatives page is a better fit.

How does Open Glean run?

Clone the repo, npm ci, npm run dev, and copy .env.example to .env.local. Set at least OPEN_GLEAN_SESSION_SECRET; everything else is optional. Open http://localhost:3000, paste a Hydra DB API key, pick a database, add an OpenRouter key in Settings, and ask away.

For a container, the README gives a docker build -t open-glean . and a docker run command with the keys and a MongoDB URI as environment variables. The image uses Next.js standalone output and runs as a non-root user. Hydra DB offers a hosted service at hydradb.com, which the maintainers run.

Who is Open Glean for?

A team that already uses Hydra DB and wants an open-source chat UI over its databases, with inline citations and deep research. Someone who wants a search over local files with no third-party service at all should look at Khoj or Onyx on the open-source alternatives to Glean page.

What limits does the README state?

Amazon DocumentDB does not support IAM database authentication, so the IAM token path cannot authenticate against it; the README says to use a standard MongoDB connection string or the Lambda proxy for DocumentDB. The optional OPEN_GLEAN_ALLOW_PRIVATE_LLM_URL setting lets a local model use plaintext HTTP on a private address; a public host still requires HTTPS, and non-http schemes are rejected. Without an OPEN_GLEAN_LLM_MODEL or a per-user model, answers are unavailable.

Questions people ask

Is Open Glean open source?

Yes, under Apache-2.0, which is an OSI-approved licence. The README notes that connector logos in the NOTICE file are third-party trademarks and are not covered by the Apache grant. The code is the Next.js server and the Hydra DB SDK integration.

Can I use Open Glean without Hydra DB?

No. The README says the Next.js server proxies requests to @hydradb/sdk, and without a Hydra key the app shows a connect screen and asks for one. Open Glean is the UI; Hydra DB is where the data and the retrieval sit. The two are built together.

Where do API keys live?

The README says Hydra and OpenRouter keys are stored in an AES-256-GCM-encrypted, httpOnly session cookie, verified server-side, and used only on the server. Nothing key-shaped is inlined into the JS bundle, localStorage or any client state. A stored key is pinned to the endpoint it was stored with, so a request may only choose the LLM base URL when it supplies its own key.

How does Open Glean compare with Glean?

Glean is a hosted enterprise search product that connects to a company's apps and runs its own retrieval. Open Glean is a self-hostable UI over Hydra DB that answers a question across memories, files and connected apps, with inline citations and a Deep Research mode. See the open-source alternatives to Glean page for how it sets alongside Khoj and Onyx.

What does Deep Research do?

The README says Deep Research is for questions one query cannot answer. The app plans a DAG of sub-questions, runs each level against Hydra in parallel, writes a finding per branch, dedupes sources into one numbered citation list, then writes the answer. A timeline shows the plan and live progress. Concurrent runs per instance default to 3 and are capped by an environment variable.

Sources

• Open Glean README and LICENSE: github.com/hydra-db/open-glean, read 2026-10-12. The README states the Apache-2.0 licence and the trademark exclusion for connector logos in NOTICE.