{"slug": "sevenbox-find-what-an-ai-agent-s-wrong-decision-affected", "title": "Sevenbox, find what an AI agent's wrong decision affected", "summary": "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.", "body_md": "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.\n\nThis repo is the open-core layer: the SDK, 14 framework adapters, the\nlocal Operations Center UI, and examples. It depends on `warrant-core`,\na separate, compiled engine package — not on PyPI, distributed as a\nGitHub Release of this repo (see [Install](#install) and\n[License](#license)).\n\nThe local Operations Center UI (`sevenbox ops`) ships with this repo —\nevery run your agent makes, with the causal graph above.\n\nNeither package is on PyPI yet. Both come from this repo.\n\n**1. Install the compiled `warrant-core` engine.** Go to this repo's\n[Releases page](https://github.com/sevenboxlabs/sevenbox/releases/tag/v0.1.1)\nand download the wheel matching your OS and Python version (3.10-3.13),\nthen:\n\n```\npip install warrant_core-0.1.1-<your-platform-and-python-tag>.whl\n```\n\nWheels are built for macOS (arm64), Linux (manylinux x86_64), and Windows\n(AMD64) — e.g. `warrant_core-0.1.1-cp312-cp312-macosx_11_0_arm64.whl` for\nmacOS on Python 3.12, or `...-manylinux_2_17_x86_64.manylinux2014_x86_64.whl`\nfor Linux, or `...-win_amd64.whl` for Windows.\n\n**2. Install `sevenbox` itself**, directly from this repo:\n\n```\npip install git+https://github.com/sevenboxlabs/sevenbox.git\n```\n\nOr, if you've already cloned it:\n\n```\npip install .\n```\n\npip resolves `sevenbox`'s `warrant-core>=0.1.1` requirement against the\nwheel you installed in step 1 and won't try to reach PyPI for it.\n\nThe first time you `import warrant_core`, you'll be shown its evaluation\nterms and asked to press Enter to accept. For CI or non-interactive use,\nset `SEVENBOX_ACCEPT_TERMS=1` instead.\n\n``` python\nfrom sevenbox_sdk import SevenboxSession, derive, irreversible, oracle, tool\n\n@oracle\ndef infer_intent(ticket):\n    return {\"intent\": \"refund_request\", \"order_id\": ticket[\"order_id\"]}\n\n@tool\ndef lookup_order(order_id):\n    return {\"status\": \"shipped\"}\n\n@derive\ndef eligibility(order):\n    return {\"eligible\": order[\"status\"] == \"shipped\"}\n\n@irreversible\n@derive\ndef refund_action(elig):\n    return {\"action\": \"refund_action(4471)\"} if elig[\"eligible\"] else None\n\nwith SevenboxSession() as session:\n    ticket = session.axiom({\"order_id\": \"4471\"})\n    action = refund_action(eligibility(lookup_order(infer_intent(ticket)[\"order_id\"])))\n    for w in session.frontier().live:\n        print(w.content)\n```\n\nSee [`docs/quickstart.md`](https://github.com/sevenboxlabs/sevenbox/blob/main/docs/quickstart.md) for the replay example and\na real-framework (LangGraph) walkthrough.\n\nValidated against 14 real agent frameworks, plus the case of running no\nframework at all (`sevenbox_sdk`'s own decorators, directly — no adapter\nneeded): LangGraph, OpenAI Agents SDK, PydanticAI, CrewAI, Deep Agents,\nMicrosoft Agent Framework, Mastra, OpenHands, LlamaIndex Agents,\nSmolagents, Letta, Google ADK, AWS Strands Agents, Haystack Agents. A\n10-15 line usage page for each lives in [`docs/adapters/`](https://github.com/sevenboxlabs/sevenbox/blob/main/docs/adapters).\n\nFinding a pending compensation obligation is half the story;\n`session.discharge()` resolves one, either because a compensating\naction was actually taken or because it turned out not to need one.\nOnce discharged, it stops showing up as pending. See\n[`examples/discharge_obligation.py`](https://github.com/sevenboxlabs/sevenbox/blob/main/examples/discharge_obligation.py)\nand [`docs/api-reference.md`](https://github.com/sevenboxlabs/sevenbox/blob/main/docs/api-reference.md) for both cases\nend to end.\n\n**222** tests in this repo, covering the SDK and all 14 adapters, pass\nagainst the real installed framework packages — no network, no API keys.\nThe full suite, including `warrant_core`/` warrant_runtime` themselves\n(private, not in this repo), is **409** tests.\n\n```\nsevenbox init    # writes a real, runnable starter script\npython my_agent.py\nsevenbox ops     # opens a local Operations Center over whatever store you point it at\n```\n\nMore at [sevenbox.dev](https://sevenbox.dev).\n\nThe SDK, adapters, and demo in this repo are licensed under\n[Apache 2.0](https://github.com/sevenboxlabs/sevenbox/blob/main/LICENSE). The `warrant-core` engine is distributed\nseparately as a compiled binary, under its own evaluation terms\n(`TERMS.md`, included with that package) — not open source, not yet\nlicensed for redistribution.", "url": "https://wpnews.pro/news/sevenbox-find-what-an-ai-agent-s-wrong-decision-affected", "canonical_source": "https://github.com/sevenboxlabs/sevenbox", "published_at": "2026-10-24 21:09:47+00:00", "updated_at": "2026-10-24 21:42:33.568682+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-safety", "artificial-intelligence"], "entities": ["Sevenbox", "Sevenbox Labs", "warrant-core", "LangGraph", "OpenAI Agents SDK", "PydanticAI", "CrewAI", "Google ADK"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/sevenbox-find-what-an-ai-agent-s-wrong-decision-affected", "markdown": "https://wpnews.pro/news/sevenbox-find-what-an-ai-agent-s-wrong-decision-affected.md", "text": "https://wpnews.pro/news/sevenbox-find-what-an-ai-agent-s-wrong-decision-affected.txt", "jsonld": "https://wpnews.pro/news/sevenbox-find-what-an-ai-agent-s-wrong-decision-affected.jsonld"}}