{"slug": "i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one", "title": "I couldn't find a simple agent framework, so I wrote one", "summary": "A developer created SwarmVault, a minimal agent framework that prioritizes simplicity over feature bloat. The framework includes 13 core skills, a shared memory vault, and an optional orchestrator for multi-agent coordination, all designed to fit in a single folder and run without background services or telemetry.", "body_md": "I keep finding a new agent skill I want to add to my library. Each one is a good addition on\n\nits own, and that is exactly the trap. Skills are not free. Every one you add takes context,\n\nand at some point it stops helping and may start making the model worse.\n\nA bigger library looks like progress. I am not convinced it is.\n\nThe frameworks I looked at had the same problem, further along. All of them were cluttered with\n\ndozens of features, a configuration surface to learn, and a cast of agents to keep straight\n\nbefore writing any code. I wanted something small enough to hold in my head.\n\nSo I combined the best of what I already knew into one set and stopped there.\n\n```\ngit clone https://github.com/AnmarHani/swarmvault && cd swarmvault && ./install.sh\n```\n\n13 is not a target. It is what was left after I took the skills I actually used or saw as good and rewrote (with the help of Claude)\n\nthem to fit together, cutting the overlap. What survived should be enough for most tasks.\n\nIf something is missing, the answer should be a thin extension, not another full skill\n\ncompeting for the same context. That constraint is the whole idea, and it is the thing I would\n\nask you to judge it on.\n\nEvery skill also runs alone. Copy one into any agent's skills directory and it works with no\n\nvault, no script, and no siblings.\n\nThe flow runs a requirements interview that ends in a real SRS, then design with the decisions\n\nwritten down, then tickets in dependency order with tests in the definition of done. The point\n\nis that the spec is not a document the agent writes and forgets. The agents follow it exactly,\n\nand at every milestone a strong model audits the work back against it.\n\nThe same standard applies to what comes out. A lot of agent output is filler: code with\n\nceremony it does not need, docs written to look complete rather than to be read. Cutting that,\n\nin code and in documentation, is part of the flow rather than something you clean up after.\n\nThe vault is a plain folder: `10 Projects`\n\n, `20 Memory`\n\n, `30 Plans`\n\n, `40 Sessions`\n\n,\n\n`50 Decisions`\n\n. The agent records what it decides is worth recording, and any later session on\n\nany platform reads it back. One zero-dependency script gives every agent `query`\n\n, `sync`\n\n,\n\n`claim`\n\n, and `context`\n\nover it.\n\nPractically, that means you can open a fresh session and say \"continue project X\" and it picks\n\nup where the last one stopped. A ticket belongs to whoever claims it first, so two agents\n\nworking in parallel do not land on the same work.\n\nOptional, off unless you turn it on. It knows which agentic platforms you have, say Claude\n\nCode and Codex and Cursor, and distributes work across them based on what each model is good\n\nat, while taking usage limits into account.\n\nThere is a central board for watching it, and it is meant to work with any CLI rather than a\n\nfixed list. Adding one is a launch-command template, not a plugin.\n\nNo dashboard, no hosted service, no plugin system. Claude Code and Codex are the verified\n\nplatforms; Gemini CLI, OpenCode, Cursor, and Copilot work from the same vault on a best-effort\n\nbasis. Linux, macOS, and WSL are first-class, native Windows is best-effort.\n\nNothing runs in the background unless you enable the orchestrator. No network calls, no\n\ntelemetry. Plain markdown on your disk, MIT.\n\nThe whole thing is small enough to read in an afternoon before you trust it with anything. That\n\nwas the point.", "url": "https://wpnews.pro/news/i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one", "canonical_source": "https://dev.to/anmarhaniv1/i-couldnt-find-a-simple-agent-framework-so-i-wrote-one-39jp", "published_at": "2026-07-26 03:43:08+00:00", "updated_at": "2026-07-26 04:29:30.373012+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "ai-tools", "ai-infrastructure"], "entities": ["SwarmVault", "AnmarHani", "Claude Code", "Codex", "Cursor", "Gemini CLI", "OpenCode", "Copilot"], "alternates": {"html": "https://wpnews.pro/news/i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one", "markdown": "https://wpnews.pro/news/i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one.md", "text": "https://wpnews.pro/news/i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one.txt", "jsonld": "https://wpnews.pro/news/i-couldn-t-find-a-simple-agent-framework-so-i-wrote-one.jsonld"}}