Guardrails AI

license

Apache-2.0(OSI) for the framework

source

public

hub

Guardrails Hub validator registry

commercial

hosted/enterprise tier

core features withheld

yes

Guardrails AI's framework and validator hub are unrestricted: one Apache-2.0 license covers the repository, and the once token-gated validator registry has moved to public PyPI. But guardrails-api, the FastAPI server that `guardrails start` launches, ships under an Elastic License adaptation that forbids offering the software to third parties as a hosted service, which is what keeps this at open core rather than open source.

• https://api.github.com/repos/guardrails-ai/guardrails/git/trees/main?recursive=1 recorded 2026-08-12

  Full untruncated recursive tree, 661 entries. One license path, the root LICENSE. No ee/, enterprise/, commercial/ or proprietary/ directory, and no license-key module. server_ci/ carries a Dockerfile, fastapi-entry.sh and compose scripts for running the server yourself.

• https://github.com/guardrails-ai/guardrails recorded 2026-06-29

  Apache-2.0 framework; ~7.1k stars; Guardrails Hub validators; commercial offerings

• https://raw.githubusercontent.com/guardrails-ai/guardrails-api/main/LICENSE recorded 2026-08-12

  The server package is not Apache-2.0. "THE FOLLOWING LICENSE HAS BEEN ADOPTED FROM THE ELSSTIC LICENSE" [sic], with the limitation "You may not provide the software to third parties as a hosted or managed service, where the service provides users with access to any substantial set of the features or functionality of the software." The repo describes itself as the "FastAPI server for hosting Guardrails AI in your own environment" and its 149-entry tree contains no license-key module, so the source is published in full - the restriction is on redistribution as a service, not on what ships.

• https://raw.githubusercontent.com/guardrails-ai/guardrails/main/docs/migration_guides/hub-to-public-pypi.md recorded 2026-08-12

  The token-gated validator registry is gone. "As of Guardrails 0.11.0, Guardrails-AI-owned validators have moved from the private Guardrails registry (pypi.guardrailsai.com, installed via guardrails hub install) to public PyPI." The change table reads "Private registry pypi.guardrailsai.com (token required)" -> "Public PyPI (no token)".

GitHub stars are the only adoption signal published for Guardrails AI: no download, install or customer figure exists for the framework. A star is not a use, so this measures interest rather than how much the framework is actually run.

• https://api.github.com/repos/guardrails-ai/guardrails recorded 2026-08-12

  stargazers_count = 7,279 for guardrails-ai/guardrails

Guardrails Hub supplies pre-built validators for toxic language, PII, prompt injection and attempts to get around safeguards, and they combine into input and output guards around a model call. The same framework also enforces structured output against a Pydantic model, so one library covers both safety checks and output format.

• https://github.com/guardrails-ai/guardrails recorded 2026-08-13

  README describes Guardrails Hub as 'a collection of pre-built measures of specific types of risks (called validators)' that 'can be combined together into Input and Output Guards that intercept' model calls, and says 'Guardrails help you generate structured data from LLMs'.

• https://guardrailsai.com/hub/ recorded 2026-09-29

  The Guardrails Hub page lists validators across risk types, including "Validates that the input string does not contain toxic language based on a BERT model.", "Detects personally identifiable information (PII) in text, using Microsoft Presidio.", "Finds prompt injection using the Rebuff prompt library." and "Detects attempts to circumvent safeguards in model conditioning."

• https://pypi.org/pypi/guardrails-ai/json recorded 2026-09-29

  The PyPI README for guardrails-ai 0.11.0 describes both jobs. Guards "detect, quantify and mitigate the presence of specific types of risks" and "Guardrails help you generate structured data from LLMs." Hub validators "can be combined together into Input and Output Guards that intercept the inputs and outputs of LLMs", and a Pydantic class can be passed so "the output is formatted to the" requested structure.

• https://www.guardrailsai.com/docs recorded 2026-09-29

  The Guardrails documentation introduction says "Guardrails is a Python framework that helps build reliable AI applications by performing two key functions", running guards "that detect, quantify and mitigate the presence of specific types of risks", and "Guardrails help you generate structured data from LLMs." Hub validators "can be combined together into Input and Output Guards that intercept the inputs and outputs of LLMs."