{
  "markdown": "# Kit memory demo · hosted MCP servers\n\nThree live, public, **read-only** MCP servers, each backed by a running\n[Kit](https://kit-project.com). Kit is an identity substrate for AI agents:\nnot another agent, not a memory API, but the durable, local-first self beneath\nyour tools, holding memory with provenance and context that outlives any\ncontext window, consolidating while it dreams. It has no chat window of its\nown; it lives under the ones you already use.\nPoint your agent at them and ask questions the model cannot answer from\ntraining. No signup; the bearer keys below are deliberately public.\n\n| server | endpoint | what it holds |\n|---|---|---|\n| Lady Whiskerdown (the Society Papers) | `https://demo-mcp.kit-project.com/mcp` | An invented regency world seen from the ballroom. 118 memories. Half of a mystery. |\n| The Area Cat (The Mews) | `https://mews-mcp.kit-project.com/mcp` | The same town from below stairs. 88 memories. The other half. |\n| The ballast case file | `https://real-mcp.kit-project.com/mcp` | A real five-year engineering decision (OpenTelemetry Collector's memory ballast), every entry with its GitHub source URL. 65 memories. |\n\nWhy fiction: none of it exists in any training corpus, so a correct citation\nis provably retrieval. Why the two archives are separate servers: the central\nmystery (who leaked the Fluffington affair?) was split across them on purpose;\nno single entry states the answer, the full corroborated case assembles only\nacross both, and your agent is the bridge. Why the real one:\nso the same machinery is checkable against the living public record.\n\n## Quickstart\n\nClaude Code:\n\n```bash\nclaude mcp add --transport http kit-papers https://demo-mcp.kit-project.com/mcp \\\n  --header \"Authorization: Bearer kbr_live_Kh9eO_ZfLldveFf1KXJEEvepfOpdgRqf7KoDDU3rB18\"\nclaude mcp add --transport http kit-mews https://mews-mcp.kit-project.com/mcp \\\n  --header \"Authorization: Bearer kbr_live_UN8-2LUlKmP2uxWQB5vpn-fzkJIuRb5zmq-apQeTU7U\"\nclaude mcp add --transport http kit-case-file https://real-mcp.kit-project.com/mcp \\\n  --header \"Authorization: Bearer kbr_live_ISWsqW39NkqT659FEkQjv0OjJDmyD4O3wx3wcCfA03c\"\n```\n\nAny MCP client: see [`examples/mcp.json`](examples/mcp.json).\n\nThen tell your agent: *\"Call wake on each server, then work out who leaked the\nFluffington affair, and cite every step.\"* Or the real-world flagship:\n*\"Should we still use the memory_ballast extension, and why did the\nrecommendation change?\"*\n\n**Keys rotate without notice.** The authoritative copy of each key is printed\nin that kit's own map: [demo llms.txt](https://demo.kit-project.com/llms.txt),\n[mews llms.txt](https://mews.kit-project.com/llms.txt),\n[real llms.txt](https://real.kit-project.com/llms.txt). If a call starts\nfailing, re-read the map.\n\n## No MCP client? No problem\n\nEvery kit also serves keyless HTTP GETs (search, read, graph pivot), documented\nin its llms.txt. A browser-only version of the whole exam runs at\n[demo.kit-project.com/investigate](https://demo.kit-project.com/investigate).\n\n## Scoped access, demonstrated\n\nThe papers kit also publishes **the apprentice key**, a deliberately narrowed\ncredential (`allowed_areas: columns`): the same kit through a smaller window.\nRun the same leak query with the full key and with the apprentice key and\nwatch the private journals vanish from the answers. That mechanism, a company\nkit sharing one knowledge area with a contractor's agent, is enforced in the\nread path, not in copy. Current apprentice key, same rotation rule:\n\n```\nkbr_live_FoAauFWl7_IoDswh8rJrFvCw2qnblQTOF8DuSBkbj7A\n```\n\n## What these servers are, honestly\n\nRead-only museum copies of Kit substrates: hybrid semantic + keyword recall,\ntyped knowledge graphs (including `supersedes` edges that keep replaced truth\ncitable), source artifacts with real URLs, and honest empty results when\nnothing is held. They demonstrate read-time behavior; a full Kit also writes,\nconsolidates while it sleeps, and hands off between sessions on your own\nmachine. There is a replayable benchmark built on these exact servers:\n[kit-memory-eval](https://github.com/PeterAtPondo/kit-memory-eval).\n\nThe build story: [kit-project.com/blog](https://kit-project.com/blog/), note\nNo. 022 in particular. The product: [kit-project.com](https://kit-project.com),\nlocal-first, invite beta.\n\n## License\n\nMIT (code and examples in this repo). The demo worlds themselves are licensed\nwith the benchmark: see kit-memory-eval's DATA-LICENSE.\n",
  "bytes": 4465,
  "sha": "c7324cbb507534bed0a2a0d6d03807e3e3b1b992b01df6420f785ebe3f191970",
  "repo_slug": "peteratpondo/kit-demo-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_peteratpondo_kit_memory_demo_7b54e7c0/readme"
}