{"slug": "claude-managed-agents", "name": "Claude Managed Agents", "category": "ai", "category_name": "AI tools and apps", "description": "", "recommenders": 3, "commenters": 7, "warners": 0, "score": 3.744, "rising": 0.35, "first": "2026-04-09", "latest": "2026-08-23", "why": "Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 3, "neutral": 3, "compare": 1, "warn": 0}, "spark": [0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "kind": "", "icon": false, "domain": "claude.com", "github_repo": null, "official_url": "", "pricing": "", "aliases": ["Claude Managed Agents"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 790498784, "item_id": 49406174, "url": "https://news.ycombinator.com/item?id=49406174", "by": "surajkrajan", "date": "2026-08-23", "version": null, "stance": "recommend", "quote": "Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "excerpt": "I have been using claude managed agents with custom rules and skills to solve open source bugs and perf problems. I analyzed 200+ repos and identified solutions. Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "context": "", "use_case": "fixing open source bugs", "switched_from": "", "story": "Using Claude hosted agents to solve open source bugs and perf improvements"}, {"id": 772184656, "item_id": 48261541, "url": "https://news.ycombinator.com/item?id=48261541", "by": "toksdotdev", "date": "2026-05-24", "version": null, "stance": "recommend", "quote": "anthropic makes this possible with self-hosted execution", "excerpt": "correct. when the sandbox is exposed via MCP, it's up to the model. the agent decides when to call it, and if it has any other way to run code (a shell tool, another MCP server), nothing forces it to go solely through the sandbox. to actually enforce isolation, the sandbox needs to be the only way the agent can run code. either don't give it any other code-running tools, or the harness exposes sandboxing as a first-c", "context": "", "use_case": "Self-hosted agent sandboxes", "switched_from": "", "story": "We made our filesystem 47\u00d7 faster by deleting it"}, {"id": 763420688, "item_id": 47713793, "url": "https://news.ycombinator.com/item?id=47713793", "by": "makeavish", "date": "2026-04-09", "version": null, "stance": "recommend", "quote": "Claude launched Managed Agents which fulfills our usecase!", "excerpt": "Claude launched Managed Agents which fulfills our usecase! https://platform.claude.com/docs/en/managed-agents/overview", "context": "building a production AI assistant", "use_case": "Production AI assistant use", "switched_from": "", "story": "Claude Agent SDK vs. Messages API for a Production AI Assistant?"}]}], "caveats": [], "who": [{"by": "makeavish", "context": "building a production AI assistant", "url": "https://news.ycombinator.com/item?id=47713793", "date": "2026-04-09"}], "tags": [], "use_cases": ["fixing open source bugs", "basic web development in vm", "agent sandbox comparison", "secrets vaults for agents", "self-hosted agent sandboxes", "production ai assistant use"], "compared": [], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=49406173", "title": "Using Claude hosted agents to solve open source bugs and perf improvements", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48849066", "title": "GPT-5.6", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48493447", "title": "Building Agents Without Harness-Engineering", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48311705", "title": "Dynamic Workflows in Claude Code", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48196831", "title": "We made our filesystem 47\u00d7 faster by deleting it", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48124443", "title": "Open Source Managed Agents", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {"recommend": 1}, "2026-05": {"compare": 1, "recommend": 1, "neutral": 1}, "2026-06": {"neutral": 1}, "2026-07": {"neutral": 1}, "2026-08": {"recommend": 1}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 790498784, "item_id": 49406174, "url": "https://news.ycombinator.com/item?id=49406174", "by": "surajkrajan", "date": "2026-08-23", "version": null, "stance": "recommend", "quote": "Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "excerpt": "I have been using claude managed agents with custom rules and skills to solve open source bugs and perf problems. I analyzed 200+ repos and identified solutions. Claude managed agents are great at running and testing code changes since the agent is provided with infra.", "context": "", "use_case": "fixing open source bugs", "switched_from": "", "story": "Using Claude hosted agents to solve open source bugs and perf improvements"}, {"id": 781642784, "item_id": 48852674, "url": "https://news.ycombinator.com/item?id=48852674", "by": "gorgmah", "date": "2026-07-09", "version": null, "stance": "neutral", "quote": "I use it for basic web stuff so haven't pushed it to the limit", "excerpt": "I use it for basic web stuff so haven't pushed it to the limit, but they have tons of stuff you can configure (up to self-hosting your environments): https://platform.claude.com/docs/en/managed-agents/environments", "context": "", "use_case": "basic web development in VM", "switched_from": "", "story": "GPT-5.6"}, {"id": 775944577, "item_id": 48496536, "url": "https://news.ycombinator.com/item?id=48496536", "by": "rajit", "date": "2026-06-11", "version": null, "stance": "neutral", "quote": "Claude Managed Agents uses one sandbox per session", "excerpt": "> what are the cost and security implications? Cost is the token usage and container uptime. > One Docker container per-customer sounds like it would be really expensive. The advantage is per-user memory and self-learning. For context, Claude Managed Agents uses one sandbox per session: https://platform.claude.com/docs/en/managed-agents/environments. > Are they started on-demand, or run 24/7? 24/7 (best for customer-", "context": "", "use_case": "agent sandbox comparison", "switched_from": "", "story": "Building Agents Without Harness-Engineering"}, {"id": 773274320, "item_id": 48329645, "url": "https://news.ycombinator.com/item?id=48329645", "by": "mrud", "date": "2026-05-29", "version": null, "stance": "neutral", "quote": "I think they have vaults[0] which provide it.", "excerpt": "I think they have vaults[0] which provide it. I don't think this is available on the non platform side. [0] https://platform.claude.com/docs/en/managed-agents/vaults", "context": "", "use_case": "secrets vaults for agents", "switched_from": "", "story": "Dynamic Workflows in Claude Code"}, {"id": 772184656, "item_id": 48261541, "url": "https://news.ycombinator.com/item?id=48261541", "by": "toksdotdev", "date": "2026-05-24", "version": null, "stance": "recommend", "quote": "anthropic makes this possible with self-hosted execution", "excerpt": "correct. when the sandbox is exposed via MCP, it's up to the model. the agent decides when to call it, and if it has any other way to run code (a shell tool, another MCP server), nothing forces it to go solely through the sandbox. to actually enforce isolation, the sandbox needs to be the only way the agent can run code. either don't give it any other code-running tools, or the harness exposes sandboxing as a first-c", "context": "", "use_case": "Self-hosted agent sandboxes", "switched_from": "", "story": "We made our filesystem 47\u00d7 faster by deleting it"}, {"id": 769991105, "item_id": 48124444, "url": "https://news.ycombinator.com/item?id=48124444", "by": "nikhil61191", "date": "2026-05-13", "version": null, "stance": "compare", "quote": "I built an open-source version of Claude-managed agents.", "excerpt": "Hello all, I built an open-source version of Claude-managed agents. The intention is to have an Agent-runtime that is model and cloud agnostic. I would love to hear feedback https://github.com/linchpinhq/linchpin", "context": "", "use_case": "", "switched_from": "", "story": "Open Source Managed Agents"}, {"id": 763420688, "item_id": 47713793, "url": "https://news.ycombinator.com/item?id=47713793", "by": "makeavish", "date": "2026-04-09", "version": null, "stance": "recommend", "quote": "Claude launched Managed Agents which fulfills our usecase!", "excerpt": "Claude launched Managed Agents which fulfills our usecase! https://platform.claude.com/docs/en/managed-agents/overview", "context": "building a production AI assistant", "use_case": "Production AI assistant use", "switched_from": "", "story": "Claude Agent SDK vs. Messages API for a Production AI Assistant?"}]}