AI Potluck
Product / UX / UI & API

PrivateGPT

Zylon

PrivateGPT is an API layer for building private AI applications against local or self-hosted models. It exposes a messages API with streaming and token counting, file and artifact ingestion, retrieval with citations and agentic RAG, built-in tools mirroring the Claude API, custom tools and MCP connectors, structured access to databases and CSV files, and embeddings. It works with any OpenAI-compatible model server.

Verified 2026-08-31 via the repository LICENSE and README.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(zylon-ai/private-gpt)
core-gated
ungated(no enterprise or ee path in the repository root and no paid tier of this code described in the README

Apache-2.0, an OSI licence, with the full source public and the shipped code in this repository. The LICENSE body was read in full: the stock Apache-2.0 text, byte-identical to the canonical template, with nothing after END OF TERMS AND CONDITIONS but the standard appendix. No enterprise or ee path in the repository root and no paid tier of this code described in the readme; zylon, the maintainer's commercial platform, is a separate end-to-end product built on top rather than a gated build of this one.

  • https://raw.githubusercontent.com/zylon-ai/private-gpt/main/LICENSE recorded 2026-08-31

    LICENSE body, read in full: the stock Apache License 2.0 text at 11,357 bytes, byte-identical to the canonical template through END OF TERMS AND CONDITIONS and followed only by the standard appendix. No appended condition.

  • https://raw.githubusercontent.com/zylon-ai/private-gpt/main/README.md recorded 2026-08-31

    README, read for the distribution and gating question. The 'What PrivateGPT gives you' section lists the messages API, ingestion, agentic RAG, built-in and custom tools, MCP connectors, database and CSV access and embeddings, all shipped from this repository. The 'PrivateGPT vs Zylon' section states that PrivateGPT is the open-source application API layer and that Zylon is a separate end-to-end platform built on top of it, adding a Triton and vLLM inference server, Kubernetes packaging and support - a product sold alongside rather than a licence-gated build of this code.

Adoption

3 low confidence
3.0

Banded on stars, because the product has no bridged distribution channel: it installs through a Homebrew tap and a shell script rather than from PyPI, and no package on PyPI belongs to it - the names `private-gpt` and `private_gpt` are unregistered, while `privategpt` and `pgpt` belong to unrelated projects. 57,491 stargazers places it at level 3, and the instrument caps the band at 3 however large the count. Stars measure attention rather than use, so confidence is low, and the honest reading is that this product's real adoption is not measurable with the instruments the routing table currently bridges.

Capability

4 medium confidence
4.0

Scored 4 on the gateway side of this category rather than the chat side. Its retrieval, tool and connector surface is the same order as anythingllm and open-webui, both 4 here, and broader than a front end like jan at 3. It stops short of 5 because it deliberately ships no interface: librechat at 5 covers multi-provider chat, multi-user authentication and agents together, where this covers the API layer alone.

  • https://raw.githubusercontent.com/zylon-ai/private-gpt/main/README.md recorded 2026-08-31

    The 'What PrivateGPT gives you' section lists a standard messages API with streaming, async and token counting, file and artifact ingestion, retrieval with citations and agentic RAG, built-in tools mirroring the Claude API, custom tools and MCP connectors, structured access to databases and CSVs, and embeddings and orchestration.

Verified 2026-08-31