Move beyond Onyx's vector search

The open-source platform that understands how your data connects, not just where the keywords match.

Self-Host for Free

→

Join the Cloud Waitlist

→

Join the Cloud Waitlist

→

Discuss Enterprise Deployment

→

Knowledge-graph-native. Open source.

Onyx is a capable open-source RAG platform, but it is fundamentally vector-native. It retrieves isolated chunks of text based on keyword proximity, and then attempts to apply an LLM to map relationships after the fact. For simple Q&A, it works. For complex enterprise workflows, it leaves you to connect the dots.

PipesHub is graph-native from day one. Instead of bolting a graph layer onto a vector database, our engine maps the relationships across your unstructured files and structured databases (SQL, warehouses) before any retrieval happens. It inherently understands that a specific Slack thread is tied to a Jira issue and a Confluence runbook.

Both platforms are open source, deploy securely in your VPC, and give you total model freedom. The difference is the engine: Pipeshub turns isolated search results into reliable, cross-app context—backed by visual citations down to the exact row or slide, and agentic workflows you can actually trust.

Features

Open-source & self-hostable

Apache 2.0

MIT

Data sovereignty (on-prem, VPC, air-gapped)

Core architecture

Graph-Native (Knowledge Graph + Vector + Agentic RAG)

Vector-Native (Vector Indexing + Agentic RAG)

Knowledge Graph role

Primary Engine. Models cross-app relationships before retrieval happens.

Post-Processing Layer. An LLM attempts to map relationships after vector search.

Data understanding

Maps cross-app relationships across files, threads, issues, and owners

Retrieves isolated text chunks by vector proximity

Citation precision

Paragraph, row, and page level with confidence levels

File-name level

Structured & unstructured data

Both

Unstructured only

Visual citations

Every document type, with confidence level

Matching text only

Model flexibility

Extensibility

Builder-centric SDK and API for custom agentic workflows

APIs and MCP server for embedding Onyx in other apps

Which Is Right

For You?

Both are open-source and self-hostable with no vendor access to your data.

The differences that matter are above the sovereignty line: how data is understood, how answers are verified, and what you can build on top.

Both platforms promise to unify your workplace data and make it searchable.

The difference is how much control, visibility, and extensibility you get once you look under the hood.

CROSS-APP CONTEXT

Teams whose questions span multiple apps, where the answer lives in the relationships between a Slack thread, a Jira ticket, and a runbook, not in any single document.

VERIFIABLE ANSWERS

Anyone who needs every answer backed by a visual citation and a confidence level, across every document type.

DEVELOPER-CENTRIC PLATFORM

Developers building proprietary, agentic workflows on a knowledge-graph backbone they can own and extend.

CONNECTED WORKFLOWS

Operations and engineering workflows over large, interconnected document sets.

STRUCTURED + UNSTRUCTURED DATA

Use cases that combine structured data (SQL, warehouses, APIs) with unstructured content.

CROSS-APP CONTEXT

Teams whose questions span multiple apps, where the answer lives in the relationships between a Slack thread, a Jira ticket, and a runbook, not in any single document.

CONNECTED WORKFLOWS

Operations and engineering workflows over large, interconnected document sets.

VERIFIABLE ANSWERS

Anyone who needs every answer backed by a visual citation and a confidence level, across every document type.

STRUCTURED + UNSTRUCTURED DATA

Use cases that combine structured data (SQL, warehouses, APIs) with unstructured content.

DEVELOPER-CENTRIC PLATFORM

Developers building proprietary, agentic workflows on a knowledge-graph backbone they can own and extend.

ENTERPRISE SEARCH

Teams whose primary need is reliable enterprise search across many tools.

SELF-SERVE PRICING

Buyers who want a transparent self-serve price and a free trial to start quickly.

MATURE & WIDELY ADOPTED

Organizations that want an established open-source project with a large community and a managed cloud option.

ENTERPRISE SEARCH

Teams whose primary need is reliable enterprise search across many tools.

MATURE & WIDELY ADOPTED

Organizations that want an established open-source project with a large community and a managed cloud option.

SELF-SERVE PRICING

Buyers who want a transparent self-serve price and a free trial to start quickly.

Frequently
Asked

Questions

Answers to the questions teams ask most when evaluating an open-source Onyx alternative.

If both are open source and both do agentic RAG, how is Pipeshub different from Onyx?

The difference is the retrieval engine. Onyx uses vector retrieval, which fetches isolated chunks of text by similarity. Pipeshub adds a knowledge graph on top of vector and agentic RAG, so it understands the relationships between pieces of content, across applications. In practice, Onyx finds the documents that mention a topic; Pipeshub connects the Slack alert, the Jira issue, and the Confluence runbook into a single contextual answer.