Sevenbox, find what an AI agent's wrong decision affected

Sevenbox Labs released Sevenbox v0.1.1, an open-core causal audit SDK for AI agent decisions that records each agent call as a "Warrant" tracking the prior claims it depended on and whether the action can be undone, then walks the causal chain to find what still needs fixing when a belief proves wrong. The release ships the SDK, 14 framework adapters (including LangGraph, OpenAI Agents SDK, PydanticAI, CrewAI and Google ADK), a local Operations Center UI and examples, and depends on the separately distributed compiled `warrant-core` engine, which is not on PyPI and must be installed from the repo's GitHub Releases page as a wheel for macOS arm64, Linux manylinux x86_64 or Windows AMD64 on Python 3.10-3.13. The repo's 222 tests, covering the SDK and all 14 adapters, pass against the real installed framework packages with no network or API keys.

Sevenbox is causal audit infrastructure for AI agent decisions. A Warrant records why a decision was made: the specific prior claims it depended on, and whether acting on it can be undone. Every call your agent makes becomes one, automatically, the moment you wrap it. When a belief your agent acted on turns out to be wrong, Sevenbox finds exactly what still needs fixing — not by searching logs, by walking the causal chain. This repo is the open-core layer: the SDK, 14 framework adapters, the local Operations Center UI, and examples. It depends on warrant-core , a separate, compiled engine package — not on PyPI, distributed as a GitHub Release of this repo see Install install and License license . The local Operations Center UI sevenbox ops ships with this repo — every run your agent makes, with the causal graph above. Neither package is on PyPI yet. Both come from this repo. 1. Install the compiled warrant-core engine. Go to this repo's Releases page https://github.com/sevenboxlabs/sevenbox/releases/tag/v0.1.1 and download the wheel matching your OS and Python version 3.10-3.13 , then: pip install warrant core-0.1.1-<your-platform-and-python-tag .whl Wheels are built for macOS arm64 , Linux manylinux x86 64 , and Windows AMD64 — e.g. warrant core-0.1.1-cp312-cp312-macosx 11 0 arm64.whl for macOS on Python 3.12, or ...-manylinux 2 17 x86 64.manylinux2014 x86 64.whl for Linux, or ...-win amd64.whl for Windows. 2. Install sevenbox itself , directly from this repo: pip install git+https://github.com/sevenboxlabs/sevenbox.git Or, if you've already cloned it: pip install . pip resolves sevenbox 's warrant-core =0.1.1 requirement against the wheel you installed in step 1 and won't try to reach PyPI for it. The first time you import warrant core , you'll be shown its evaluation terms and asked to press Enter to accept. For CI or non-interactive use, set SEVENBOX ACCEPT TERMS=1 instead. python from sevenbox sdk import SevenboxSession, derive, irreversible, oracle, tool @oracle def infer intent ticket : return {"intent": "refund request", "order id": ticket "order id" } @tool def lookup order order id : return {"status": "shipped"} @derive def eligibility order : return {"eligible": order "status" == "shipped"} @irreversible @derive def refund action elig : return {"action": "refund action 4471 "} if elig "eligible" else None with SevenboxSession as session: ticket = session.axiom {"order id": "4471"} action = refund action eligibility lookup order infer intent ticket "order id" for w in session.frontier .live: print w.content See docs/quickstart.md https://github.com/sevenboxlabs/sevenbox/blob/main/docs/quickstart.md for the replay example and a real-framework LangGraph walkthrough. Validated against 14 real agent frameworks, plus the case of running no framework at all sevenbox sdk 's own decorators, directly — no adapter needed : LangGraph, OpenAI Agents SDK, PydanticAI, CrewAI, Deep Agents, Microsoft Agent Framework, Mastra, OpenHands, LlamaIndex Agents, Smolagents, Letta, Google ADK, AWS Strands Agents, Haystack Agents. A 10-15 line usage page for each lives in docs/adapters/ https://github.com/sevenboxlabs/sevenbox/blob/main/docs/adapters . Finding a pending compensation obligation is half the story; session.discharge resolves one, either because a compensating action was actually taken or because it turned out not to need one. Once discharged, it stops showing up as pending. See examples/discharge obligation.py https://github.com/sevenboxlabs/sevenbox/blob/main/examples/discharge obligation.py and docs/api-reference.md https://github.com/sevenboxlabs/sevenbox/blob/main/docs/api-reference.md for both cases end to end. 222 tests in this repo, covering the SDK and all 14 adapters, pass against the real installed framework packages — no network, no API keys. The full suite, including warrant core / warrant runtime themselves private, not in this repo , is 409 tests. sevenbox init writes a real, runnable starter script python my agent.py sevenbox ops opens a local Operations Center over whatever store you point it at More at sevenbox.dev https://sevenbox.dev . The SDK, adapters, and demo in this repo are licensed under Apache 2.0 https://github.com/sevenboxlabs/sevenbox/blob/main/LICENSE . The warrant-core engine is distributed separately as a compiled binary, under its own evaluation terms TERMS.md , included with that package — not open source, not yet licensed for redistribution.