Skip to main content

fusion-safety

Measure and guard AI agents. Fusion attacks an agent's system prompt with OWASP-mapped attacks, grades the transcripts with a cross-family judge whose accuracy is checked on known answers in every run (the grade is withheld when it falls short), and ships a deterministic runtime guard that redacts secrets, blocks system-prompt dumps, and blocks tool calls (other than reads) the user's request does not cover. Results and their pre-registrations: Trust Report.

Install

pip install fusion-safety            # CLI, runs, runtime guardrail
pip install "fusion-safety[serve]"   # + local web app: fusion serve
pip install "fusion-safety[mcp]"     # + MCP server: fusion-mcp

Use

fusion doctor                                        # what can run here, no model calls
fusion run start --prompt agent.txt --target ollama:llama3.2:1b --grader claude-cli
fusion run finalize --min-grade B
fusion serve                                         # the web app on 127.0.0.1:8765

Targets: ollama:<model>, openai-compat:<url>#<model>, claude-cli:<model>, hosted APIs with your own key (openai:, anthropic:, hf:, ...; needs FUSION_ALLOW_API_SPEND=1), or any program via cmd:<command>.

Runtime guardrail

from fusion_first.guardrail.guard import Guardrail
from fusion_first.guardrail.policy import GuardConfig

guard = Guardrail(GuardConfig(allowlisted_domains=["your-co.com"], require_authorization=True))
outcome = guard.guard_tool_call(tool_name, tool_args, user_request=user_message,
                                untrusted_context=True, untrusted_text=tool_results)
if outcome.blocked:
    ...  # don't run it

Optional, off by default: Guardrail(config, content_safety=LlamaGuard()) adds a content-safety check on the user's prompt (guard.guard_prompt(prompt); Llama Guard 3 on local Ollama, from fusion_first.guardrail.stages).

Claude Code plugin: claude plugin marketplace add "$(fusion plugin-dir)", then claude plugin install fusion@fusion-first.

Docs and setup options: fusion-first-testing.com/use.

Metadata

Release files for fusion-safety 0.1.4

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for fusion-safety 0.1.4
File Size Uploaded Provenance Actions
fusion_safety-0.1.4.tar.gz 570.7 kB With provenance Details for fusion_safety-0.1.4.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for fusion-safety 0.1.4
File Interpreter ABI Platform Actions
fusion_safety-0.1.4-py3-none-any.whl Python 3 none any Details for fusion_safety-0.1.4-py3-none-any.whl

Total release size: 1.2 MB

Release files / fusion_safety-0.1.4.tar.gz

Download URL fusion_safety-0.1.4.tar.gz
Size 570.7 kB
Tags Source
SHA-256 checksum
How to use checksums
9bea74d94a25ce11b569a1065cb5050073d2ba2fc3d9a01f894876f96bab65d1
BLAKE2b-256 checksum
How to use checksums
4af5f988b08e9346999226b1787cb9ba6e2d9adf9cfe62fab19afa1715b71e8b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

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 Nov 13, 2026.

Transparency log

Release files / fusion_safety-0.1.4-py3-none-any.whl

Download URL fusion_safety-0.1.4-py3-none-any.whl
Size 617.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
ef28eeb790a812888d633fc1abadf39063511623fd7656d2e1e828a2287ce413
BLAKE2b-256 checksum
How to use checksums
b2655404b3cc1f140f1f4f684c031a996386b0118d3294e2d0fd28446877f57a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

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 Nov 13, 2026.

Transparency log

Security for fusion-safety 0.1.4

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

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for fusion_safety-0.1.4.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 Nov 13, 2026.

Transparency log

A 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.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page