{"slug": "cedar", "name": "Cedar", "category": "other", "category_name": "Other software", "description": "", "recommenders": 3, "commenters": 5, "warners": 0, "score": 3.093, "rising": -0.98, "first": "2025-12-23", "latest": "2026-03-18", "why": "Instead of LLM as a judge, we use the Cedar policy language to authorize tool calls.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 4, "neutral": 2, "compare": 0, "warn": 0}, "spark": [0, 0, 0, 1, 3, 0, 0, 0, 0, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "Instead of LLM as a judge, we use the Cedar policy language to authorize tool calls.", "kind": "", "icon": true, "domain": null, "github_repo": null, "official_url": "https://www.cedarpolicy.com/", "pricing": "", "aliases": ["Cedar"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 758479249, "item_id": 47404953, "url": "https://news.ycombinator.com/item?id=47404953", "by": "aaguiarz", "date": "2026-03-16", "version": null, "stance": "recommend", "quote": "you are better suited with a different solution that does not require a server roundtrip like Cedar", "excerpt": "Actually, you don't _need_ to sync data, you can send it contextually too https://openfga.dev/docs/interacting/contextual-tuples. Of course, if you end up sending all data contextually, you are better suited with a different solution that does not require a server roundtrip like Cedar. This repo compares differences between the OpenFGA approach and the Cedar approach https://github.com/openfga/openfga-cedar-compariso", "context": "", "use_case": "", "switched_from": "", "story": "Event Publisher enables event integration between Keycloak and OpenFGA"}, {"id": 757164049, "item_id": 47322753, "url": "https://news.ycombinator.com/item?id=47322753", "by": "joshdevon", "date": "2026-03-10", "version": null, "stance": "recommend", "quote": "Instead of LLM as a judge, we use the Cedar policy language to authorize tool calls.", "excerpt": "The last six months of agent security have been a dumpster fire of vulnerabilities\u2014from EchoLeak to CurXecute. Most current security relies on \"prompt and pray\" or sandboxing. Prompts are suggestions, not constraints. Sandboxes are a blunt instrument that often break agent utility. At Unprompted conference, we released our work on a deterministic Reference Monitor for coding agents like Claude Code, Cursor, and GitHu", "context": "", "use_case": "authorizing agent tool calls", "switched_from": "", "story": "Hooking Coding Agents with the Cedar Policy Language"}, {"id": 756587569, "item_id": 47286723, "url": "https://news.ycombinator.com/item?id=47286723", "by": "pjmlp", "date": "2026-03-07", "version": null, "stance": "recommend", "quote": "Cedar, a full graphics workstation in the evolution of Mesa", "excerpt": "Including how to build a full workstation with memory safe systems languages, besides Smalltalk, Interlisp-D, now recovered and playable in the browser, https://interlisp.org/ Mesa with XDE (think the safety Zig is selling today, but in 1985) http://bitsavers.informatik.uni-stuttgart.de/pdf//xerox/xde/XDE_3.0_Nov84/610E00130_XDE_Concepts_and_Principles_Sep85.pdf Cedar, a full graphics workstation in the evolution of ", "context": "", "use_case": "", "switched_from": "", "story": "Smalltalk's Browser: Unbeatable, yet Not Enough"}, {"id": 750216369, "item_id": 46888523, "url": "https://news.ycombinator.com/item?id=46888523", "by": "joshdevon", "date": "2026-02-04", "version": null, "stance": "recommend", "quote": "creates deterministic lanes for the agent using Cedar (AWS's policy as code language)", "excerpt": "As we all know, OpenClaw is awesome precisely because it gives us Simon Willison\u2019s lethal trifecta: Access to private data, exposure to untrusted content, and the ability to externally communicate. While extremely risky, it gives us a glimpse of the future we can have if we actually could trust agents. To date, sandboxing (or buying mac-minis) has been the approach to reducing risk. While necessary, sandboxes also ma", "context": "built a policy extension for OpenClaw", "use_case": "policy enforcement for agent tool calls", "switched_from": "", "story": "Show HN: We told OpenClaw to rm -rf and it failed successfully"}]}], "caveats": [], "who": [{"by": "joshdevon", "context": "built a policy extension for OpenClaw", "url": "https://news.ycombinator.com/item?id=46888523", "date": "2026-02-04"}], "tags": [], "use_cases": ["authorizing agent tool calls", "policy enforcement for agent tool calls", "authorization policy language"], "compared": [{"slug": "lean", "name": "Lean", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47407996", "title": "item 47407996", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47399437", "title": "Event Publisher enables event integration between Keycloak and OpenFGA", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47322752", "title": "Hooking Coding Agents with the Cedar Policy Language", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47258885", "title": "Smalltalk's Browser: Unbeatable, yet Not Enough", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46888509", "title": "Show HN: We told OpenClaw to rm -rf and it failed successfully", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46364973", "title": "Test, don't just verify", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {"neutral": 1}, "2026-01": {}, "2026-02": {"recommend": 1}, "2026-03": {"recommend": 3, "neutral": 1}, "2026-04": {}, "2026-05": {}, "2026-06": {}, "2026-07": {}, "2026-08": {}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 758762448, "item_id": 47422653, "url": "https://news.ycombinator.com/item?id=47422653", "by": "d_christiansen", "date": "2026-03-18", "version": null, "stance": "neutral", "quote": "Cedar (https://lean-lang.org/use-cases/cedar/) at AWS use an executable Lean model", "excerpt": "Cedar (https://lean-lang.org/use-cases/cedar/) at AWS use an executable Lean model of a system as an oracle for differential testing of a Rust implementation. If you can't run Lean in production, then their approach is compelling. The idea is that you start with a Lean specification, create a fully verified implementation with respect to the spec, and then hook it and the production implementation up to a fuzzer or s", "context": "", "use_case": "", "switched_from": "", "story": null}, {"id": 758479249, "item_id": 47404953, "url": "https://news.ycombinator.com/item?id=47404953", "by": "aaguiarz", "date": "2026-03-16", "version": null, "stance": "recommend", "quote": "you are better suited with a different solution that does not require a server roundtrip like Cedar", "excerpt": "Actually, you don't _need_ to sync data, you can send it contextually too https://openfga.dev/docs/interacting/contextual-tuples. Of course, if you end up sending all data contextually, you are better suited with a different solution that does not require a server roundtrip like Cedar. This repo compares differences between the OpenFGA approach and the Cedar approach https://github.com/openfga/openfga-cedar-compariso", "context": "", "use_case": "", "switched_from": "", "story": "Event Publisher enables event integration between Keycloak and OpenFGA"}, {"id": 757164049, "item_id": 47322753, "url": "https://news.ycombinator.com/item?id=47322753", "by": "joshdevon", "date": "2026-03-10", "version": null, "stance": "recommend", "quote": "Instead of LLM as a judge, we use the Cedar policy language to authorize tool calls.", "excerpt": "The last six months of agent security have been a dumpster fire of vulnerabilities\u2014from EchoLeak to CurXecute. Most current security relies on \"prompt and pray\" or sandboxing. Prompts are suggestions, not constraints. Sandboxes are a blunt instrument that often break agent utility. At Unprompted conference, we released our work on a deterministic Reference Monitor for coding agents like Claude Code, Cursor, and GitHu", "context": "", "use_case": "authorizing agent tool calls", "switched_from": "", "story": "Hooking Coding Agents with the Cedar Policy Language"}, {"id": 756587569, "item_id": 47286723, "url": "https://news.ycombinator.com/item?id=47286723", "by": "pjmlp", "date": "2026-03-07", "version": null, "stance": "recommend", "quote": "Cedar, a full graphics workstation in the evolution of Mesa", "excerpt": "Including how to build a full workstation with memory safe systems languages, besides Smalltalk, Interlisp-D, now recovered and playable in the browser, https://interlisp.org/ Mesa with XDE (think the safety Zig is selling today, but in 1985) http://bitsavers.informatik.uni-stuttgart.de/pdf//xerox/xde/XDE_3.0_Nov84/610E00130_XDE_Concepts_and_Principles_Sep85.pdf Cedar, a full graphics workstation in the evolution of ", "context": "", "use_case": "", "switched_from": "", "story": "Smalltalk's Browser: Unbeatable, yet Not Enough"}, {"id": 750216369, "item_id": 46888523, "url": "https://news.ycombinator.com/item?id=46888523", "by": "joshdevon", "date": "2026-02-04", "version": null, "stance": "recommend", "quote": "creates deterministic lanes for the agent using Cedar (AWS's policy as code language)", "excerpt": "As we all know, OpenClaw is awesome precisely because it gives us Simon Willison\u2019s lethal trifecta: Access to private data, exposure to untrusted content, and the ability to externally communicate. While extremely risky, it gives us a glimpse of the future we can have if we actually could trust agents. To date, sandboxing (or buying mac-minis) has been the approach to reducing risk. While necessary, sandboxes also ma", "context": "built a policy extension for OpenClaw", "use_case": "policy enforcement for agent tool calls", "switched_from": "", "story": "Show HN: We told OpenClaw to rm -rf and it failed successfully"}, {"id": 741866336, "item_id": 46366646, "url": "https://news.ycombinator.com/item?id=46366646", "by": "loglog", "date": "2025-12-23", "version": null, "stance": "neutral", "quote": "how their existing system (Cedar) works", "excerpt": "The author is in the comfortable position of working on a system that does have a formal specification and a formally verified reference implementation. The post is not about how they wish things would work, but how their existing system (Cedar) works. Regarding your point on Rust, the vast majority of software has nowhere near the amount of static guarantees provided by Rust. If you need more, use static memory allo", "context": "", "use_case": "Authorization policy language", "switched_from": "", "story": "Test, don't just verify"}]}