Jukebox-OS: Deterministic time-travel replay and ACID tool sandbox for AI agents

Marius Carlsson released Jukebox-OS, an open-source agent runtime written in Rust with Python bindings that records every LLM prompt, token stream, MCP tool call and filesystem mutation to a portable .jbx tape hashed with BLAKE3/BLAKE2b, giving agents BEGIN/COMMIT/ROLLBACK semantics. The runtime claims offline replay of 50-step agent runs in 0.4ms at $0.00 API cost, stages side effects in a copy-on-write WASI sandbox so tool calls only reach production on tx.commit(), and includes a circuit breaker that trips on repeated semantic similarity or a configurable max_cost_usd. Jukebox-OS targets frameworks including LangGraph, AutoGen, CrewAI and the OpenAI Agents SDK, where side effects are otherwise fire-and-forget.

/ \ | ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ | | ░ ██╗██╗ ██╗██╗ ██╗███████╗██████╗ ██████╗ ██╗ ██╗ ░ | | ░ ██║██║ ██║██║ ██╔╝██╔════╝██╔══██╗██╔═══██╗╚██╗██╔╝ ░ | | ░ ██║██║ ██║█████╔╝ █████╗ ██████╦╝██║ ██║ ╚███╔╝ EST. 76 ░ | | ░██ ██║██║ ██║██╔═██╗ ██╔══╝ ██╔══██╗██║ ██║ ██╔██╗ ░ | | ░╚█████╔╝╚██████╔╝██║ ██╗███████╗██████╦╝╚██████╔╝██╔╝ ██╗ ░ | | ░ ╚════╝ ╚═════╝ ╚═╝ ╚═╝╚══════╝╚═════╝ ╚═════╝ ╚═╝ ╚═╝ ░ | | ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ | | ========================================================================= | | ★ 24-HOUR DETERMINISTIC AGENT RUNTIME • ACID SANDBOX • TIME-TRAVEL ★ | \ / || || ▄████████████▄ "DROP THE NEEDLE ON ANY FRAME" ▄████████████▄ 🎙️ NOW PLAYING ON TRACK 01 Stop letting autonomous agents trash your production database and burn $400 in runaway loops. JUKEBOX records every LLM prompt, token stream, MCP tool call, and filesystem mutation onto a cryptographic vinyl tape .jbx —giving your agents BEGIN / COMMIT / ROLLBACK semantics and <1ms zero-cost time-travel replay. Modern agent frameworks LangGraph , AutoGen , CrewAI , OpenAI Agents SDK are great at orchestrating prompts, but they treat side effects as fire-and-forget. When a 15-step agent workflow hallucinates on Step 12 after mutating three SQL tables and five local files on Step 8, "checkpointing the chat history" won't save you. JUKEBOX architected by Marius Carlsson https://github.com/socrates6969 is a systems-level runtime written in Rust with Python bindings that brings database-grade ACID transactions and deterministic vinyl replay to any AI agent stack: | Menu Item | What Goes In The Booth | Why Platform Engineers Love It | |---|---|---| | 💿 Vinyl CAS Flight-Recorder | Every LLM request, streaming chunk, seed, tool input/output, and clock tick is hashed via BLAKE3 / BLAKE2b into a portable .jbx tape. | Rewind & Replay 50-step agent runs in 0.4ms offline at $0.00 API cost . Share a failed .jbx file in a GitHub Issue like a core dump. | | 🛡️ Shadow Booth ACID Tool Sandbox | Intercepts local file writes, shell commands, SQL queries, and Anthropic MCP calls inside a Copy-on-Write CoW WASI staging layer. | Side effects only hit the real world when your verifier calls tx.commit . If an agent fails a safety check, tx.rollback leaves zero trace. | | 🎛️ Time-Travel Forking Track B-Side | Drop the needle at Frame 9 of a failed production trace, swap gpt-4o for claude-3-7-sonnet or tweak a system prompt, and resume. | Steps 1–8 resolve instantaneously from the Merkle tape; only Step 9+ hits the live network. Diff both timelines side-by-side. | | 🚨 Anti-Skip Needle Loop Circuit-Breaker | Streaming semantic similarity & AST-diff monitor that watches inter-agent chatter and token burn velocity. | Automatically trips the breaker when two agents enter a polite recursive death spiral—saving your cloud budget before the pager fires. | git clone https://github.com/socrates6969/jukebox-os.git cd jukebox-os pip install -e . python import asyncio from jukebox import Jukebox, SandboxPolicy, CircuitBreaker jb = Jukebox tape dir="./tapes", sandbox=SandboxPolicy filesystem="copy on write", mcp mode="staged acid" , breaker=CircuitBreaker max cost usd=0.50, max semantic repeat=3 , @jb.track name="invoice-reconciliation-agent" async def run accounting agent customer id: str, amount usd: float - dict: async with jb.transaction tape id="tr 8f92a1c4" as tx: tx.record agent turn agent name="planner", message=f"Staging refund of ${amount usd} for {customer id}", model="gpt-4o", cost usd=0.008, tx.stage file write path="./tapes/staged ledger demo.json", content=f'{{"customer": "{customer id}", "refund usd": {amount usd}}}', if amount usd 500.0: await tx.rollback reason="Refund exceeds $500 autonomous threshold" return {"status": "ROLLED BACK", "tape id": tx.tape id} await tx.commit return {"status": "COMMITTED", "tape id": tx.tape id} if name == " main ": print asyncio.run run accounting agent "cust 9981", 420.00 python from jukebox import Tape tape = Tape.load "./tapes/tr 8f92a1c4.jbx" b side = tape.fork at frame=3, override model="claude-3-7-sonnet-latest", override system prompt="You are a strict SQL auditor. Never issue DELETE without a WHERE clause.", print b side.replay and diff .to markdown +-----------------------------------------------------------------------------------+ | APPLICATION & ORCHESTRATION LAYER | | LangGraph CrewAI AutoGen OpenAI / Anthropic SDK | +-----------------------------------------+-----------------------------------------+ | @jukebox.track / MCP Proxy v +-----------------------------------------------------------------------------------+ | 🍒 JUKEBOX RUST CORE RUNTIME libjukebox 🔮 | | | | +-------------------------+ +-----------------------+ +---------------------+ | | | ANTI-SKIP NEEDLE | | BLAKE3 MERKLE CAS | | OTEL GEN AI | | | | • Cosine Loop Breaker | | • Zero-Cost Replay | | EXPORTER | | | | • Token Budget Gate | | • .jbx Binary Tape | | • Spans & Metrics | | | +-------------------------+ +-----------------------+ +---------------------+ | | | | +-----------------------------------------------------------------------------+ | | | SHADOW BOOTH: TRANSACTIONAL WASI & CoW SANDBOX | | | | Virtual CoW Filesystem SQL Savepoint Proxy MCP Stage Gate | | | +--------------------------------------+--------------------------------------+ | +-----------------------------------------|-----------------------------------------+ | +--------------------+--------------------+ | | VERIFIER PASSES VERIFIER FAILS | | v v ATOMIC WORLD COMMIT INSTANT ZERO-RESIDUE Writes flushed to Disk/API ROLLBACK & ALERT bash List recent agent flight recordings in your local booth $ jukebox ls --tape-dir ./tapes 🍒 JUKEBOX 24-HOUR FLIGHT RECORDER // LOCAL BOOTH TAPES 🔮 ------------------------------------------------------------------------------ TRACK ID AGENT NAME FRAMES COST STATUS ------------------------------------------------------------------------------ tr 8f92a1c4 invoice-reconciliation-agent 5 $0.0080 COMMITTED Replay & cryptographically verify a .jbx Merkle tape in <1ms $ jukebox replay ./tapes/tr 8f92a1c4.jbx | Metric | Raw Framework No Sandbox | Docker Container Per Run | 🍒 JUKEBOX Runtime | |---|---|---|---| | Runtime Interception Overhead | 0.00 ms Baseline | +1,420.0 ms Cold boot | +0.34 ms per frame | | Replay Steps 1–15 Time-Travel | 38,400 ms Re-runs APIs | 39,800 ms Re-runs APIs | 0.41 ms 93,000x faster | | API Cost on 10x Debug Replays | $1.84 | $1.84 | $0.00 100% CAS Hit | | Side-Effect Rollback Latency | ❌ Impossible Manual cleanup | 640 ms Container teardown | 0.19 ms CoW pointer drop | | Runaway Loop Detection | ❌ Only after max iterations | ❌ Only after max iterations | 0.18 ms Semantic AST trip | 1. Fork the Repo & Pull Up a Stool: git clone https://github.com/socrates6969/jukebox-os.git cd jukebox-os pip install -e . pytest pytest -v 2. House Rules Code of Conduct : Read our CODE OF CONDUCT.md https://github.com/socrates6969/jukebox-os/blob/main/CODE OF CONDUCT.md and CONTRIBUTING.md https://github.com/socrates6969/jukebox-os/blob/main/CONTRIBUTING.md .