Osprey Framework
🎉 Latest Release: v2026.6.2 - Open & self-hosted model support: run the agent on any OpenAI-protocol model as a configuration change, with a new model-capability benchmark; plus facility-timezone unification, ARIEL publishing fixes, and hardening
🚧 Early Access Release This is an early access version of the Osprey Framework. While the core functionality is stable and ready for experimentation, documentation and APIs may still evolve. We welcome feedback and contributions!
A production-ready framework for deploying agentic AI in large-scale, safety-critical control system environments—particle accelerators, fusion experiments, beamlines, and complex scientific facilities.
📄 Research This work was presented as a contributed oral presentation at ICALEPCS'25 and will be featured at the Machine Learning and the Physical Sciences Workshop at NeurIPS 2025.
🚀 Quick Start
# Install the framework as a standalone CLI tool (using uv, recommended)
uv tool install osprey-framework
# Create a minimal project to verify your setup
osprey build quickstart --preset hello-world
cd quickstart
# If API keys aren't already in your environment, copy and edit .env:
# cp .env.example .env
# Start a Claude Code agent session
claude
For a production project tailored to your detector, beamline, or accelerator subsystem, install the guided osprey-build-interview skill and run it from Claude Code:
# Install the /osprey-build-interview skill into ~/.claude/skills/
uv run osprey skills install osprey-build-interview
Then start Claude Code in an empty directory and type /osprey-build-interview. The
skill walks you through a guided conversation, produces a build profile, and
osprey build profile.yml generates a ready-to-use project.
📚 Documentation
📖 Read the Full Documentation →
🧪 Testing
# Run unit tests (fast, no API keys required)
pytest tests/ --ignore=tests/e2e -v
# Run e2e tests (slow, requires API keys)
# ⚠️ IMPORTANT: Use 'pytest tests/e2e/' NOT 'pytest -m e2e'
pytest tests/e2e/ -v
See tests/e2e/README.md and the Contributing Guide for details.
Key Features
- Agent-Driven Orchestration - Skills, MCP tools, and explicit dependency declarations let the Osprey agent decompose operator requests into auditable steps with mandatory approval gates
- Control-System Safety - Pattern detection, PV boundary checking, and mandatory approval for hardware writes
- Protocol-Agnostic Integration - Seamless connection to EPICS, LabVIEW, Tango, and mock environments
- Scalable Capability Management - Dynamic classification prevents prompt explosion as toolsets grow
- Production-Proven - Deployed at major facilities including LBNL's Advanced Light Source accelerator
📖 Citation
If you use the Osprey Framework in your research or projects, please cite our paper:
@article{10.1063/5.0306302,
author = {Hellert, Thorsten and Montenegro, João and Sulc, Antonin},
title = {Osprey: Production-ready agentic AI for safety-critical control systems},
journal = {APL Machine Learning},
volume = {4},
number = {1},
pages = {016103},
year = {2026},
month = {02},
doi = {10.1063/5.0306302},
url = {https://doi.org/10.1063/5.0306302},
}
For detailed installation instructions, tutorials, and API reference, please visit our complete documentation.
Copyright Notice
Osprey Framework Copyright (c) 2025, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.
If you have questions about your rights to use or distribute this software, please contact Berkeley Lab's Intellectual Property Office at IPO@lbl.gov.
NOTICE. This Software was developed under funding from the U.S. Department of Energy and the U.S. Government consequently retains certain rights. As such, the U.S. Government has been granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, distribute copies to the public, prepare derivative works, and perform publicly and display publicly, and to permit others to do so.
Release files for osprey-framework 2026.6.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| osprey_framework-2026.6.2.tar.gz | 15.3 MB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| osprey_framework-2026.6.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 18.1 MB
Release files / osprey_framework-2026.6.2.tar.gz
| Download URL | osprey_framework-2026.6.2.tar.gz |
|---|---|
| Size | 15.3 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
00c125357308116d694628d54cf332ec7834c060d15a8cf0cec4297eb82b7b90
|
|
BLAKE2b-256 checksum How to use checksums |
09f72f7fb20ca070070309e9f6bd690cbdfaf28e8111ee2f6de22e70068ff044
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 28, 2026.
Transparency logRelease files / osprey_framework-2026.6.2-py3-none-any.whl
| Download URL | osprey_framework-2026.6.2-py3-none-any.whl |
|---|---|
| Size | 2.8 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0d9af62297e279baadd8ab92e7720f14908b9fb1ea22119d0c830d73acdd9fc3
|
|
BLAKE2b-256 checksum How to use checksums |
58faca8c9eed75f7d6b75793d46e57f277ef4a1eb03d253cb7230d053fd5c6c5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 28, 2026.
Transparency logSecurity for osprey-framework 2026.6.2
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for osprey_framework-2026.6.2.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 28, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
