AI Sandboxes
Overview
AI sandboxes are controlled execution environments where agents can run code, browse, inspect files, call tools, or manipulate artifacts without putting the host system at unnecessary risk. A sandbox gives the agent enough power to do real work while limiting filesystem, network, credential, and process access.
Conference Context
The pattern comes from operating-system isolation, browser sandboxes, CI runners, notebooks, container platforms, and secure code-execution services. Agentic coding and computer-use systems made sandboxing a default requirement rather than a specialty feature.
How This Theme Evolved
- World's Fair 2024 (local comparison fixture): Human oversight and infrastructure were broad safeguards around AI systems.
- Miami 2026 (public comparison wiki): Remote agents and sandboxed compute made runtime boundaries part of developer workflow.
- World's Fair 2025 (local comparison fixture): The comparison corpus foregrounded containment, identity, OAuth, access control, and code-execution safety.
- World's Fair 2026 (current event synthesis backed by linked local evidence): The linked conference graph treats sandboxes, tool permissions, provenance, and eval gates as one operational control surface.
Confidence: high. Boundary: Earlier event wikis are comparison context only. They are not primary evidence for World's Fair 2026 claims, and no local fixture is represented as a public live site. Comparison source: aie wiki generation delta.
Significance
Agents need to experiment, test, and inspect state. Sandboxes let them do that while containing failures, malicious inputs, runaway processes, and accidental destructive changes.
Why This Matters Now
Agents increasingly execute code and use credentials rather than only returning text, so correctness and blast-radius control must be designed together.
WF26 evidence gate: this section was emitted only because the page links to configured local evidence. Relevant configured evidence: 2026 06 30 robert brennan sandboxes aren t optional runtime isolation patterns for coding agents at scale, 2026 06 30 samuel colvin your agent needs a sandbox not a desert.
Confidence: high. Comparison history is context; linked WF26 pages remain the evidence for current-event claims.
Applied Use
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Choose isolation based on risk: separate processes for low-risk tasks, containers or microVMs for untrusted code, and policy-controlled network and secret access for production work. Capture logs, diffs, artifacts, and resource usage so human operators can review what happened.
They are useful in coding assistants, data-analysis agents, browser agents, app builders, test runners, educational tools, and any system that executes generated code or commands.
Use a sandbox whenever an agent can execute code, inspect user files, download dependencies, browse unknown sites, or run untrusted scripts. Loosen limits only after the workflow and threat model are well understood.
Practical Lesson
Give each run the minimum tools, credentials, network reach, filesystem scope, and lifetime it needs; capture actions and outputs so approval and incident review use the same evidence trail.
Confidence: high. Treat this as synthesis derived from the linked evidence graph, not as an official schedule claim.
Connections
- 2026 06 30 pierluca d oro computer use at the edge of the statistical precipice — Computer Use at the Edge of the Statistical Precipice; Pierluca D'Oro (Day 3 — Session Day 2 · 11:10am-11:30am · Computer Use; official schedule)
- 2026 06 30 robert brennan sandboxes aren t optional runtime isolation patterns for coding agents at scale — Sandboxes Aren't Optional: Runtime Isolation Patterns for Coding Agents at Scale; Robert Brennan (Day 3 — Session Day 2 · 3:20pm-3:40pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 30 samuel colvin your agent needs a sandbox not a desert — Your agent needs a sandbox, not a desert; Samuel Colvin (Day 3 — Session Day 2 · 12:05pm-12:25pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 29 tushar jain unlock agent autonomy the runtime for ai native systems — Unlock Agent Autonomy: The Runtime for AI-Native Systems; Tushar Jain (Day 2 — Session Day 1 · 3:45pm-4:05pm · AI Architects: Show my Workflow; official schedule)
- 2026 06 30 abhishek bhardwaj from fork to fleet designing an agent sandbox cloud pt 1 — From fork() to Fleet: Designing an Agent Sandbox Cloud Pt 1; Abhishek Bhardwaj (Day 3 — Session Day 2 · 1:30pm-1:50pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 30 abhishek bhardwaj from fork to fleet designing an agent sandbox cloud pt2 — From fork() to Fleet: Designing an Agent Sandbox Cloud Pt2; Abhishek Bhardwaj (Day 3 — Session Day 2 · 1:55pm-2:15pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 30 ivan burazin kubernetes is not your sandbox — Kubernetes Is Not Your Sandbox; Ivan Burazin (Day 3 — Session Day 2 · 11:40am-12:00pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 30 kevin orellana 1 000 agent tasks in a sandbox what breaks when llms write and run code — 1,000 Agent Tasks in a Sandbox: What Breaks When LLMs Write and Run Code; Kevin Orellana (Day 3 — Session Day 2 · 2:25pm-2:45pm · Sandbox & Platform Engineering; official schedule)
- 2026 06 30 adam azzam don t build agents build environments — Don’t build agents, build environments; Adam Azzam (Day 3 — Session Day 2 · 10:45am-11:05am · Sandbox & Platform Engineering; official schedule)
- 2026 06 29 matt brockman how i learned to stop worrying and love the sandbox — How I learned to stop worrying and love the sandbox; Matt Brockman (Day 1 — Workshop Day · 11:05am-12:05pm · Workshops Day 1; official schedule)
- 2026 06 29 alexander embiricos the golden age of ai engineering — The Golden Age of AI Engineering; Alexander Embiricos, Romain Huet (Day 2 — Session Day 1 · 9:25am-9:45am · Software Factories; verified event YouTube resource; via youtube pMggiOb18tc)
- 2026 06 30 liad yosef mcp apps extending the frontier — MCP Apps - Extending the frontier; Liad Yosef, Ido Salomon (Day 3 — Session Day 2 · 2:25pm-2:45pm · Context Engineering; verified event YouTube resource; via youtube o zkvb0iFDQ)
- 2026 07 01 arun sekhar blast radius zero one command openclaw sandboxes in the cloud — Blast Radius Zero: One‑Command OpenClaw Sandboxes in the Cloud; Arun Sekhar (Day 4 — Session Day 3 · 1:55pm-2:15pm · Track M; official schedule)
- 2026 06 29 derek meegan deploying browser agents at scale — Deploying browser agents at scale; Derek Meegan (Day 2 — Session Day 1 · 1:55pm-2:15pm · Expo Stage 4 SE; official schedule)
- 2026 06 29 ross taylor scaling to long horizons algorithms environments compute — Scaling to Long-Horizons: Algorithms, Environments, Compute; Ross Taylor, Chengxi Taylor (Day 2 — Session Day 1 · 2:25pm-2:45pm · Data Quality; official schedule)
- 2026 07 01 miguel gonz lez fern ndez the art of building verifiers for computer use agents — The Art of Building Verifiers for Computer Use Agents; Miguel González Fernández, Corby Rosset (Day 4 — Session Day 3 · 11:40am-12:00pm · Expo Stage 1 NE; official schedule)
- 2026 07 01 rowan christmas yolo mode safely microvm sandboxes for any agent — YOLO Mode, Safely: microVM Sandboxes for Any Agent; Rowan Christmas (Day 4 — Session Day 3 · 1:30pm-1:50pm · Expo Stage 2 NW; official schedule)
- 2026 06 29 rayan garg rethinking environments for long horizon work — Rethinking Environments for Long Horizon Work; Rayan Garg (Day 2 — Session Day 1 · 11:40am-12:00pm · Data Quality; official schedule)
- 2026 06 29 mahesh sathiamoorthy data and environment curation for post training llms — Data and Environment Curation for Post-training LLMs; Mahesh Sathiamoorthy (Day 2 — Session Day 1 · 3:45pm-4:05pm · Data Quality; official schedule)
- 2026 06 30 tina manghnani from framework to runtime running agents with foundry agent service — From framework to runtime: running agents with Foundry Agent Service; Tina Manghnani, Keiji Kanazawa (Day 3 — Session Day 2 · 10:45am-11:05am · Track M; official schedule)
- 2026 06 30 viren baraiya harnessing agents the durable runtime for dynamic workflows — Harnessing Agents: The Durable Runtime for Dynamic Workflows; Viren Baraiya (Day 3 — Session Day 2 · 11:10am-11:30am · Expo Stage 1 NE; official schedule)
- 2026 07 01 yohei nakajima active graph agent runtime babyagi 4 — Active Graph Agent Runtime (BabyAGI 4); Yohei Nakajima (Day 4 — Session Day 3 · 11:10am-11:30am · Graphs; official schedule)
- 2026 06 29 ang li the autonomous computer full stack infrastructure for computer use agents — The Autonomous Computer: Full-stack Infrastructure for Computer Use Agents; Ang Li (Day 1 — Workshop Day · 4:30pm-5:30pm · Workshops Day 1; official schedule)
- 2026 06 30 philipp schmid why agents should have their own sandbox — Why Agents Should Have Their Own Sandbox; Philipp Schmid (Day 3 — Session Day 2 · 1:30pm-1:50pm · Expo Stage 1; official schedule)
- John Craft
- Abhishek Bhardwaj
- Arun Sekhar
- Will Brown
- Tina Manghnani
- Pierluca D'Oro
- Robert Brennan
- Samuel Colvin
- Tushar Jain
- Ivan Burazin
- Kevin Orellana
- Adam Azzam
- Matt Brockman
- Alexander Embiricos
- Romain Huet
- Liad Yosef
- Ido Salomon
- Derek Meegan
- Ross Taylor
- Chengxi Taylor
- Miguel González Fernández
- Corby Rosset
- Rowan Christmas
- Rayan Garg
- Docker
- Microsoft
- OpenAI
- Browserbase
- Amazon AGI Lab
- Prime Intellect
- MCP Apps
- Cua
- typedef
- Oxylabs
- Amazon Web Services
- Navan
- Uber
- Warp
- Meta
- Yugabyte
- Programma Labs
- OpenHands
- 2026 06 29 alexander embiricos the golden age of ai engineering — The Golden Age of AI Engineering; Alexander Embiricos, Romain Huet (Day 2 — Session Day 1 · 9:25am-9:45am · Software Factories; related YouTube resource; via youtube pMggiOb18tc)
- 2026 06 30 liad yosef mcp apps extending the frontier — MCP Apps - Extending the frontier; Liad Yosef, Ido Salomon (Day 3 — Session Day 2 · 2:25pm-2:45pm · Context Engineering; related YouTube resource; via youtube o zkvb0iFDQ)
- youtube 4kYl2_mqmnQ slides — I Run a Fleet of AI Agents Across Three Machines. Here's What Broke. - Kyle Jaejun Lee, KRAFTON (10 extracted slide frames)
- youtube qdZzND79mcg slides — Beyond the Harness: A Journey Towards Adaptative Engineering - Rajiv Chandegra, Annicha Labs (16 extracted slide frames)
- youtube 2IxD9OB3XuQ slides — Continual Learning for AI Agents: From Failures to Durable Improvements - Soheil Feizi, RELAI (24 extracted slide frames)
- youtube sAOBXCDiDOs slides — MCP Apps: Primitives, discovery, and the Future of Software - Pietro Zullo, Manufact, Inc (18 extracted slide frames)
- youtube vljxQZfJ9wY slides — Production Evals For Agentic AI Systems - Nishant Gupta, Meta Superintelligence Labs (12 extracted slide frames)
- 2026 06 29 du an lightfoot agents that own their inference building production ai agents on dedicated gpus — Agents That Own Their Inference: Building Production AI Agents on Dedicated GPUs; Du'an Lightfoot (Day 1 — Workshop Day · 9:00am-11:00am · Track 7; related YouTube resource; via youtube wFTVEDYVJT0)
- 2026 06 30 paul klein iv bringing agents onto the world wide web — Bringing agents onto the world wide web; Paul Klein IV (Day 3 — Session Day 2 · 11:40am-12:00pm · Computer Use; official schedule)
Evidence Graph
Transcript-backed resources
- youtube V EDrhIhHzQ — The Prime Intellect Stack — Will Brown, Prime Intellect
- youtube ZSQb5fzRFPw — Computer-Use 2.0: Agents Just Got Multi-Cursor — Francesco Bonacci, Cua
- youtube pMggiOb18tc — The Golden Age of AI Engineering — Alexander Embiricos & Romain Huet & Peter Steinberger, OpenAI
- youtube o zkvb0iFDQ — MCP UI: Extending the frontier — Liad Yosef and Ido Salomon, MCP Apps
Quote signals
- “So there is a lot of like dark magic like happening behind the wood just to uh give you some context.” — youtube ZSQb5fzRFPw
- “And if you ask us how we trust that agent, the answer is that it's just evows all the way down.” — youtube ZSQb5fzRFPw
Transcript-backed resources
Transcript-backed resources
Transcript-backed resources
Transcript-backed resources
- youtube Rx8f05JI_WA — SWE-Marathon: Evaluating Coding Agents at Billion-Token Scale - Rishi Desai, Abundant AI
- youtube 2IxD9OB3XuQ — Continual Learning for AI Agents: From Failures to Durable Improvements - Soheil Feizi, RELAI
- youtube SKDJo2CopRs — Why Eval++ Is the Next Great Compute Primitive — Sunil Pai & Matt Carey, Cloudflare
- youtube JnubYCYunk8 — Browser Agents Don't Need Better Models. They Need Better Eyes. - Kushan Raj, ARK
- youtube wFTVEDYVJT0 — Building Agents with Amazon Nova Act and MCP - Du'An Lightfoot, Amazon (Full Workshop)
- youtube c 2eEv2ou7Y — Why MCP and ChatGPT Apps Use Double Iframes — Frédéric Barthelet, Alpic
- youtube TNwJ1LMiENk — Stop Making Models Bigger, Make Them Behave — Kobie Crawford, Snorkel
- youtube grdoOC1BT1s — Think You Can Build a Game with AI? Think Again! - Danielle An & David Hoe, Meta
- youtube 4uFVSLgD2Q4 — Agents in Production: How OpenGov Built and Scaled OG Assist - Gabe De Mesa, OpenGov
- youtube YYH0DMQr30A — Task Fidelity Scaling Laws — Kobie Crawdord, Snorkel
- youtube UPwGaM2MKHY — The Log Is The Agent - Ishaan Sehgal, Omnara
- youtube bRnoEpoK5m4 — The Pipeline Is Dead - Iris ten Teije, Sky Valley Ambient Computing
- youtube ILdE7FaAjVA — Under 5 minutes to a deployed LLM endpoint — Audry Hsu, RunPod
- youtube _B4Pv9ttFgY — Building Agent Interfaces: Lessons from Chrome DevTools (MCP) for Agents — Michael Hablich, Google
- youtube 4kYl2_mqmnQ — I Run a Fleet of AI Agents Across Three Machines. Here's What Broke. - Kyle Jaejun Lee, KRAFTON
- youtube qlHaO6laBlM — Shipping Production AI Inside Government — William Tarr, Ministry of Justice
- youtube ghJmWQCIHRM — The agent-ready web: Simplify user actions with WebMCP — Tara Agyemang, Google
- youtube HvZXAOZ3iv8 — What Lies Beneath the API — Benjamin Cowen, Modal
Transcript-backed resources
This evidence graph consolidates scheduled talks, linked videos, transcripts, and slide-derived material connected to this topic.
Linked Sessions
- Computer Use at the Edge of the Statistical Precipice
- Sandboxes Aren't Optional: Runtime Isolation Patterns for Coding Agents at Scale
- Your agent needs a sandbox, not a desert
- Unlock Agent Autonomy: The Runtime for AI-Native Systems
- 'From fork() to Fleet: Designing an Agent Sandbox Cloud Pt 1'
- 'From fork() to Fleet: Designing an Agent Sandbox Cloud Pt2'
- Kubernetes Is Not Your Sandbox
- 1,000 Agent Tasks in a Sandbox: What Breaks When LLMs Write and Run Code
- Don’t build agents, build environments
- How I learned to stop worrying and love the sandbox
Media Signals
youtube-rcsliSIy_YU— 8 slide-derived text signals- Slide-derived themes for
youtube-rcsliSIy_YU: coding, code, snippets, generation, single, atomic, tasks, curl. - Evidence links for
youtube-rcsliSIy_YU: youtube rcsliSIy_YU, youtube rcsliSIy_YU slides, youtube rcsliSIy_YU dense slides, youtube rcsliSIy_YU reconstructed slides youtube-bmWZk9vTze0— 10 slide-derived text signals- Slide-derived themes for
youtube-bmWZk9vTze0: query, client, info, table, tool, call, response, await. - Evidence links for
youtube-bmWZk9vTze0: youtube bmWZk9vTze0, youtube bmWZk9vTze0 slides, youtube bmWZk9vTze0 dense slides, youtube bmWZk9vTze0 reconstructed slides youtube-wsFd22SL1s8— 10 slide-derived text signals- Slide-derived themes for
youtube-wsFd22SL1s8: systems, chrome, code, sandboxes, operating, distributed, windows, subsystem. - Evidence links for
youtube-wsFd22SL1s8: youtube wsFd22SL1s8, youtube wsFd22SL1s8 slides, youtube wsFd22SL1s8 dense slides, youtube wsFd22SL1s8 reconstructed slides youtube-e9sLVMN76qU— 8 slide-derived text signals- Slide-derived themes for
youtube-e9sLVMN76qU: most, today, tooling, breaks, moment, remove, human, loop. - Evidence links for
youtube-e9sLVMN76qU: youtube e9sLVMN76qU, youtube e9sLVMN76qU slides, youtube e9sLVMN76qU reconstructed slides
Source Coverage
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
| Evidence type | Count | Review note |
| --- | ---: | --- |
| other | 40 | Related pages outside the main evidence categories. |
| resources | 22 | Video/resource pages; check source status before treating as primary event evidence. |
| slides | 16 | OCR or reconstructed slide evidence; mark claims as OCR-derived unless image-reviewed. |
| talks | 24 | Official schedule pages; use for titles, speakers, tracks, and stated talk framing. |
| tools | 5 | Derived inventory pages; use as entity context, not independent proof. |
Talks
- 2026 06 30 pierluca d oro computer use at the edge of the statistical precipice
- 2026 06 30 robert brennan sandboxes aren t optional runtime isolation patterns for coding agents at scale
- 2026 06 30 samuel colvin your agent needs a sandbox not a desert
- 2026 06 29 tushar jain unlock agent autonomy the runtime for ai native systems
- 2026 06 30 abhishek bhardwaj from fork to fleet designing an agent sandbox cloud pt 1
- 2026 06 30 abhishek bhardwaj from fork to fleet designing an agent sandbox cloud pt2
Resources
- youtube wFTVEDYVJT0
- youtube Rx8f05JI_WA
- youtube 2IxD9OB3XuQ
- youtube SKDJo2CopRs
- youtube JnubYCYunk8
- youtube c 2eEv2ou7Y
Slides
- youtube 4kYl2_mqmnQ slides
- youtube qdZzND79mcg slides
- youtube 2IxD9OB3XuQ slides
- youtube sAOBXCDiDOs slides
- youtube vljxQZfJ9wY slides
- youtube rcsliSIy_YU slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
| other | 44 | Related pages outside the main evidence categories. |
| resources | 24 | Video/resource pages; check source status before treating as primary event evidence. |
| talks | 26 | Official schedule pages; use for titles, speakers, tracks, and stated talk framing. |
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
| resources | 25 | Video/resource pages; check source status before treating as primary event evidence. |
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
| other | 45 | Related pages outside the main evidence categories. |
| resources | 26 | Video/resource pages; check source status before treating as primary event evidence. |
| slides | 17 | OCR or reconstructed slide evidence; mark claims as OCR-derived unless image-reviewed. |
Talks
Resources
Slides
Tools
This table summarizes the local evidence already linked from this topic. It is a navigation aid, not a claim that every linked page has been fully reviewed.
| resources | 27 | Video/resource pages; check source status before treating as primary event evidence. |
Talks
Resources
Slides
Tools
Active Use Cases
- Running generated code and tests before suggesting a patch.
- Browser or computer-use automation with constrained state.
- Temporary workspaces for data analysis and document transformation.
- Reproducible agent task environments for evaluations.
Slide-Derived Supporting Decks
- youtube CnA2lGfymY slides — "I've never seen anything scarier than an LLM with tool calls." — Erik Meijer aka @HeadinTheBox (32 extracted slide frames)
- youtube ZSQb5fzRFPw slides — (17 extracted slide frames)
These decks are slide/OCR support only; keep the article synopsis, origin, use cases, and schedule sections as the primary topic narrative.
Slide-Derived Scheduled Session Signals
- 2026 06 29 erik meijer in code they act in proof we trust — In Code They Act, In Proof We Trust