{"slug": "show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab", "title": "Show HN: Knowl – agent memory with write-time supersession, 0.90 on MAB", "summary": "Knowl, an agent memory tool that runs as an MCP server, announced it scored 90 out of 100 on the MemoryAgentBench FactConsolidation-SH benchmark at 262K, addressing the problem of stale agent memory. The tool, version 5.14.0, records reasoning, evidence, and provenance for every fact, and supersedes outdated information without deleting history, with integrations for Claude, Codex, Cursor, Copilot, and others.", "body_md": "### Checkable by design\n\nEvery fact carries its reasoning, its evidence and its provenance — observed, stated by a person, or inferred. A claim you can check without asking anybody.\n\nReasoning, evidence and provenance on every fact\n\nBecause it gets trusted. Knowl gives every fact your agents store **a record of where it came from**, a **review trail**, and a **successor when it changes** — so the answer they hand back is the current one, and you can see who settled it.\n\n[How this was built](https://blog.knowl.cloud/the-story-of-knowl) — the approaches that failed first, and the measurements that ruled each one out.\n\n~/your-repo\n\n``` bash\n$ npm install -g @dat999zx/knowl\nadded 1 package\n$ knowl --version\n5.14.0\n```\n\nNode 20 or newer. Nothing else to install.\n\nA session ends and the context goes with it.\n\nIt gets re-proposed. **Six months on, nobody can say why it was decided.**\n\nSomebody acts on it. **Stale knowledge is worse than none, because it gets trusted.**\n\nThe next one starts from zero. **The same ground gets relearned every week.**\n\nEvery fact carries its reasoning, its evidence and its provenance — observed, stated by a person, or inferred. A claim you can check without asking anybody.\n\nReasoning, evidence and provenance on every fact\n\nKnowl runs in your repository and keeps its knowledge there. Your code, your embeddings and your session history are never uploaded.\n\nLocal by default, shared on purpose\n\nA fact records what it replaced and what replaced it. Superseded knowledge is never deleted — there is no delete verb in the product at all.\n\nHistory is the feature, not a changelog\n\nAdapters\n\nKnowl runs as an MCP server, so anything that speaks MCP works with no adapter to write. Three of these also take lifecycle hooks, which is what lets an agent capture what it learned without being asked.\n\n`knowl init claude`\n\nMCP + hooks`knowl init codex`\n\nMCP + hooks`knowl init cursor`\n\nMCP + hooks`knowl init copilot`\n\nMCP + hooks`knowl init openhands`\n\nMCP + hooks`knowl init antigravity`\n\nMCP + hooks`knowl init windsurf`\n\nMCP + gate`knowl init cline`\n\nMCP + plugin`knowl acp`\n\nMCP + ACP`knowl acp`\n\nMCP + ACP`knowl acp`\n\nMCP + ACP`knowl acp`\n\nMCP + ACP`knowl init claude-desktop`\n\nMCP`knowl init opencode`\n\nMCP`speaks MCP`\n\nbring your ownBefore / after\n\nMost teams already have agent memory. It is a markdown file nobody trusts, and a person who remembers the rest.\n\nFreshness\n\nA memory tool that only stores things goes stale silently. This is what happens to a single claim over four months — every step recorded, nothing overwritten, nothing deleted.\n\n“What's our rate limit on the public API?”\n\n100 requests a minute*fact**observed** api-gateway*\n\nApproved in the review queue\n\n“What's our rate limit on the public API?”\n\n500 requests a minute*supersedes the March fact*\n\nThe March answer is still there\n\nProof\n\nWhen a fact changes, most memory tools keep handing back the old one. On the public benchmark for exactly this, Knowl got 90 of 100 right.\n\nSubEM on [MemoryAgentBench](https://github.com/HUST-AI-HYZ/MemoryAgentBench) FactConsolidation-SH at 262K, scored with the benchmark’s own harness. Method, raw results, and the splits where we do worse: [the full write-up](https://github.com/dat999zx/knowl/blob/main/benchmarks/memoryagentbench/mab/FINDINGS.md).\n\nIn their words\n\nMemory going stale is not our framing of the problem. It is written into the documentation of the agents your team already runs.\n\nAnthropic\n\n“Over many sessions a memory store accumulates duplicates, contradictions, and stale entries.”\n\nManaged Agents documentation — [Dreams](https://platform.claude.com/docs/en/managed-agents/dreams)\n\nOpenAI\n\n“Memory can become stale. Agents are instructed to treat memories as guidance only.”\n\nAgents SDK — [agent memory](https://openai.github.io/openai-agents-python/sandbox/memory/)\n\nGreyhound Research\n\n“The deepest flaw in current agent memory is that it mistakes retrieval for memory.”\n\nSanchit Vir Gogia, Chief Analyst — [via InfoWorld](https://www.infoworld.com/article/4191031/microsoft-unveils-memora-to-tackle-ai-agents-memory-problem.html)\n\nQuoted verbatim from published sources, read August 2026. These are observations about the problem, not endorsements of Knowl — every link goes to the page the sentence was read on.\n\nPricing\n\nQueries never reach us — your agents read a replica on your own machine, so reading back is free at every tier and always will be.\n\nAccounts are free and so is your first workspace. Anyone can be invited in and use all of it — you pay when your team outgrows what Free carries each period.\n\nA hosted memory layer for coding agents. Your agents write what they work out to a local store; publishing lifts a fact into a shared workspace your team reads from.\n\nNo. The index, the embeddings and every session stay on your machine. Publishing is an explicit act, and only the fact you publish travels.\n\nIt is superseded, not deleted. The new fact records what it replaced, the old one stays queryable as history, and only the current answer is returned by default.\n\nA wiki has no idea when it went stale, and a vector store returns the old claim beside the new one. Knowl records supersession, so the answer you get is the current one.\n\nClaude Code, Cursor, Codex, Gemini CLI and Claude Desktop, configured by one command. Anything that speaks MCP works without an adapter.\n\nOnly the members of the workspace you published into. Knowledge written privately stays on your machine until someone promotes it deliberately.\n\nEvery write is secret-scanned on the server, on the way in, every time — not once at setup. A refused write names what tripped it without echoing the value back.\n\nOne repository is enough. Connect it, sign the CLI in, and let a week of work accumulate before you judge it.", "url": "https://wpnews.pro/news/show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab", "canonical_source": "https://knowl.cloud", "published_at": "2026-08-27 14:06:26+00:00", "updated_at": "2026-08-27 14:19:01.957126+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools"], "entities": ["Knowl", "MemoryAgentBench", "Anthropic", "OpenAI", "Greyhound Research", "Sanchit Vir Gogia"], "alternates": {"html": "https://wpnews.pro/news/show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab", "markdown": "https://wpnews.pro/news/show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab.md", "text": "https://wpnews.pro/news/show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab.txt", "jsonld": "https://wpnews.pro/news/show-hn-knowl-agent-memory-with-write-time-supersession-0-90-on-mab.jsonld"}}