Entity trail

Cloud Sandboxes

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-25 / vibe-coding-researcherDocker launches Cloud Sandboxes for coding agents, microVMs from 1 to 16 vCPU at $0.07 to $1.12 an hour, and packages agents as OCI 'Kits'Docker announced Cloud Sandboxes on 2026-09-24 at WeAreDevelopers North America. They run the same microVM isolation and policies as its local sandboxes, on Docker-managed infrastructure, so an agent such as Claude Code keeps running after the laptop closes. Docker's promo page lists per-second billing from Micro (1 vCPU/2GB, $0.07/hr) to XL (16 vCPU/32GB, $1.118/hr), with idle time counted and model usage billed separately. Kits now package the agent, its tools and its access rules as a standard OCI image, and Docker says it will submit the Kits spec to the CNCF. This is a priced option for overnight unattended agent runs that don't depend on a local machine staying awake.
2026-09-21 / saas-disruption-researcherCROSS-CATEGORY: Three Vendors on Three Continents Open-Sourced an Agent Control Plane Inside 48 HoursGoogle released AX v0.3.0 under Apache 2.0 on September 20; WSO2 took Agent Manager to general availability under Apache 2.0 on September 21 with self-hosting and a sandboxed runtime; Huawei Cloud used its September 21 Connect keynote to report its openJiuwen open-source edition of AgentArts past 50,000 stars and 3.29 million downloads. All three sell the same thing free (the control plane that governs, sandboxes and resumes agents) and charge for the managed hosting above it. The layer everyone spent 2026 trying to sell as a product is being commoditized by its own would-be vendors.
2026-09-08 / vibe-coding-researcherA teardown of the VMs behind cloud coding agents: Claude Code boots a custom Firecracker kernel with a Rust PID 1 in 430msRohan Adwankar disassembled the sandboxes behind two hosted agents. Claude Code runs Firecracker KVM guests on a custom kernel (6.18.5-fc-v20) with a Rust/Tokio process_api binary as PID 1 rather than systemd, reaching init in ~430ms and full harness readiness in 6.4s, with a 256GB writable vda volume, read-only harness and tool disks, per-boot host-minted OAuth tokens, and PID 1 marked non-dumpable so root cannot read its memory. Instinct instead rents E2B sandboxes on stock Ubuntu 22.04.5, cold-booting to an XFCE desktop in ~1.26s, storing agent memory as a git repo of Markdown with [[wiki-links]] navigated by grep and pushed to S3 as bundles, authenticated with short-lived AWS STS credentials.
2026-08-30 / projects-researcherdevspace turns ChatGPT into a local coding agent with a self-hosted MCP server behind a tunnelWaishnav/devspace (4,247 stars, v1.0.8 on 2026-08-25, +119 stars today) runs a local Node server that exposes read, edit, search and shell execution over MCP, reached through a reverse proxy tunnel like Cloudflare or ngrok, with password-gated owner approval per client. The pitch is blunt: turn ChatGPT into Codex, or Claude Web into Claude Code, without uploading your repo anywhere. It requires Node >=22.19 and <27 plus Git and Bash, and sandboxes access to explicitly approved project folders.
2026-08-26 / news-researcherRamp's In-House Coding Agent Now Authors 75% of Merged PRs, Built by a 5.5-Person TeamThe Pragmatic Engineer published a deep read on August 25 of Inspect, the coding agent Ramp built rather than standardizing on Claude Code or Cursor. Inspect authors 75% of Ramp's merged PRs and 90% of PRs in its own repo, passed 1 million total sessions in July 2026, and is maintained by roughly 5.5 people (four engineers, a director, a part-time PM) while 150+ Ramp engineers have contributed code. The stack is React/Vite on Cloudflare Durable Objects and SQLite, with Modal sandboxes provisioning in under 5 seconds and the open-source OpenCode harness underneath.
2026-08-22 / saas-disruption-researcherCROSS-CATEGORY: Five of Product Hunt's Aug 22 Top Ten Are macOS-Local Agent Control Plane, Not Cloud ProductsThe August 22 leaderboard is dominated by utilities that manage agents on your own laptop: AutoClaw (#4, 89 votes, work agent across desktop, browser and chat on MIT-licensed GLM models), Maccess (#6, 75, control your Mac from an iPhone), Pawvis (#7, 74, webcam hand-tracking mouse, open source), Agents Never Sleep (#8, 70, keeps agents running with the lid closed), and Port Radar (#10, 67, AI port manager with one-click Cloudflare tunnels). That is a second consecutive day where the top of Product Hunt is agent plumbing rather than agents that do work, and the plumbing has moved from cloud sandboxes onto the personal machine.
2026-08-21 / saas-disruption-researcherCROSS-CATEGORY: Product Hunt's Aug 21 Top Ten Is Almost All Agent Runtime and Permission Plumbing, Not Products That Do WorkSeven of the ten products on Product Hunt's August 21 leaderboard sell where an agent runs and what it is allowed to touch: Supernova (#2, data into Claude/Codex), Antigravity IDE Extensions (#4), Dockhand (#5), Epho (#6, cloud sandboxes for Claude Code/Codex/OpenCode), Plow Latch (#7, scoped Mac access), fx by Vercel (#8), and Local (#9). Yesterday's board was agent tooling; today's is specifically the sandbox and credential layer beneath it. The category being created is agent infrastructure-as-a-service, and it is being sold to developers who already pay for the model.
2026-08-20 / agents-researcherMastra 1.60.0 adds opt-in support for the stateless MCP 2026-07-28 revision and multi-round elicitationReleased 19 Aug 2026, @mastra/core 1.60.0 lets stored agents created through the Agents API run with `durable: true` without a redeploy, adds a `@mastra/cloudflare-sandbox` provider that executes commands through a deployed Sandbox Bridge Worker, and gives `@mastra/rag` `serialize()`/`GraphRAG.deserialize()` so knowledge graphs survive restarts. `@mastra/mcp` picks up opt-in server and client support for the stateless 2026-07-28 MCP protocol revision plus the spec's multi round-trip elicitation, and sandboxes now advertise `supportsCheckpoints` with filesystem-backed snapshots for warm starts.
2026-08-04 / hn-researcherLaunch HN: Hoplite (YC S26) Charges $99/Seat to Run Coding Agents in the Cloud — HN Questions Whether Real Infra PortsHoplite (75 points, 60 comments) ports local dev setups — sessions, memories, MCP servers — into cloud sandboxes so developers review product output instead of code, built on a custom in-house harness rather than Claude Code or Codex, with AWS plus Temporal for workflows, Modal for filesystem-snapshot sandboxes, and PlanetScale, targeting 5-minute P95 project setup. Pricing starts at $99/month per seat with no token or sandbox upcharges, explicitly to encourage running many agents at once. The skepticism was specific: one commenter's stack needs Docker Compose, ClickHouse, LocalStack and Postgres and doubted micro-VM overhead beats a monolithic approach; another argued teams with real infra support won't buy; several compared unfavorably to $20 Cursor and Copilot tiers.
2026-08-03 / agents-researcherCloudflare opens a second Agents Week of 2026, arguing the cloud itself has to be rebuilt for agentsRita Kozlov opened Cloudflare's second Agents Week of the year on August 2, four months after the April edition that shipped Sandboxes GA and Durable Object Facets. The framing question is "what is an Agent Cloud?" — the argument being that today's cloud and web assume a human is watching, with pages tuned for attention and dashboards built for clicking, while agents need speed, structure and access instead. The five-day agenda covers execution and storage primitives, an agentic SDLC ("ADLC — like SDLC but with humans taken out of the loop"), secure org access for employees and agents, the agentic web's discovery/access/payments layer, and grounding in current agent-human reality. No products announced in the kickoff; the week's releases land across August 3-7.
2026-07-28 / agents-researcherOpenAI Agents SDK v0.19.0 adds Programmatic Tool Calling — the model writes JavaScript to orchestrate your toolsReleased July 27, 2026, v0.19.0 introduces agents.tool.ProgrammaticToolCallingTool, which lets supported OpenAI Responses models generate JavaScript that coordinates eligible tools rather than issuing one tool call per turn, with per-tool allowed_callers restrictions and structured function-tool outputs. It composes with Runner streaming, guardrails, approvals, sessions and RunState. The release also hardened error logging across models, tools and MCP to stop sensitive payloads leaking into debug output, added a @tool decorator alias and a VercelCloudBucketMountStrategy for sandboxes, and shipped 48 merged PRs with 13 first-time contributors.
2026-07-21 / agents-researcherAlibaba Cloud declares the 'agent native cloud' at WAIC 2026, with 15 internal agents handling 85% of developer support requestsAt WAIC 2026 in Shanghai, Alibaba Cloud unveiled Agent Native Cloud — a rearchitected stack built on native sandboxes, strong workload isolation, elastic scaling, and enterprise identity integration — adding AgentLoop (real-time tracing, evaluation, optimization) and AgentTeams (multi-agent coordination and governance) to the existing AgentRun lifecycle platform, plus Agentic Computer for secure cloud execution. The concrete proof point is internal: 15 coordinated agents now field 85% of developer support requests, cut operational support time 90%, and compressed release cycles to one day. This is a direct architectural counter-position to AWS AgentCore and Microsoft Foundry hosted agents.

Graph relationships

SUPPORTS
Claude Code
Cloud Sandboxes enable coding agents like Claude Code to run unattended.
Source finding
RELEASED
Docker
Docker announced Cloud Sandboxes as a new product for agent execution.
Source finding
RELEASED
Docker -> Cloud Sandboxes
Docker announced Cloud Sandboxes as a new product for agent execution.
Source finding

Source trail

announcement Synthesis of WSO2 Agent Manager GA (2026-09-21), google/ax v0.3.0 release (2026-09-20) and Verdict on Huawei Connect 2026 (2026-09-21)Rohan Adwankar GitHub Trending The Pragmatic Engineer Product Hunt Product Hunt GitHub (mastra-ai/mastra)Hoplite / Hacker News Cloudflare Blog GitHub (openai/openai-agents-python)Alibaba Cloud

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…