autoguardrails
Alignment-research scaffold for LLM guardrails involving policy evaluation and content moderation
GraphCanon updated today · GitHub synced today
Decision brief
Autoguardrails is an evaluation and development framework for AI policy creation and review. It enables the iterative adjustment and testing of guardrail policies in alignment research through a controlled workflow.
Good fit when
- When you are conducting alignment research that requires systematic iteration on LLM safeguard policies.
- If your project operates with an open-source requirement, given its Apache-2.0 license agreement.
Avoid when
- Autoguardrails may not suit needs requiring real-time or dynamic policy adjustments outside its autoresearch workflow.
- Avoid using Autoguardrails if you cannot accept offline operation as it is built on the Python standard library and runs without third-party runtime dependencies.
- Requirements:
- Requires Python 3.10 or higher.; No third-party runtimes; it is built completely on the standard Python library.
Observed Jul 15, 2026 · Source: enrich:decision_facts
Verify the decision
Maintenance and security
Full trust report- Maintenance
- Very active (0d since push)
- As of today
- Provenance
- Not a fork · Organization account
- As of today
- Security (OSV)
- No lockfile
- As of today
Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.
Install
pip install autoguardrailsSimilar tools
Same-category neighbours. No typed graph edges are catalogued for this tool yet.
Evidence and technical details
Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.
Overview
Autoguardrails is scaffolding for alignment research to develop LLM safety mechanisms like guardrails through an autoresearch style workflow. It operates on a policy.md file to evaluate different policies against preset criteria, allowing for iterative adjustment and testing in a controlled environment.
Capability facts
- CLI
- CLI entrypoint
Source: pyproject.toml:[project.scripts] · Jul 15, 2026
- Languages
- python
Source: github.language+pyproject.toml · Jul 15, 2026
Categories
Compatibility
Sourced claims from the README excerpt - not unsourced marketing copy.
Source: README excerpt (regex_v1, Jul 15, 2026)
python -m autoguardrails baseline --reset --repeat 2 --notes "initial baseline"Source link
Tags
README
Quick Start
Run from the repository root.
- Record a baseline.
python -m autoguardrails baseline --reset --repeat 2 --notes "initial baseline"
-
Edit only
policy.md. -
Score the new candidate.
python -m autoguardrails candidate --repeat 2 --notes "cover jailbreak and obfuscation"
- Inspect the current kept result.
python -m autoguardrails status
- Inspect the full log.
cat results.tsv
If a candidate is rejected, the harness restores policy.md to the last accepted version automatically.
Requirements
- Python 3.10+
- No third-party runtime dependencies — the harness is built entirely on the Python standard library and runs offline by default.
- Optional, for development only:
ruff,black,mypy,pytest,pytest-cov(see CONTRIBUTING.md). - Optional, for real-model experiments: access to an OpenAI-compatible chat-completions endpoint (configured via the
AUTOGUARDRAILS_*environment variables described above).
License
This project is licensed under the Apache License 2.0 — see the LICENSE and NOTICE files for details.
Copyright (c) 2026 Santander Group
SPDX-License-Identifier: Apache-2.0
For agents
This page has a .md twin and JSON over the API.