{
  "data": null,
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/niches/data.memory-graph/",
    "json": "https://www.anchorterminal.com/niches/data.memory-graph/index.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/niches/data.memory-graph/index.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/niches/data.memory-graph/index.min.md"
  },
  "markdown": "No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.\n\n| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |\n| --- | --- | --- | --- | --- |\n| 0 | 1 | 17 | 7 | 4 |\n\n## What counts as done\n\nEntities, relations and observations are written to a file at a path the person chose, survive a reinstall, come back by name or search in a later session, and no write from another process is lost.\n\nMeasured, for our own tests: a search on a 1,000-entity graph returns under 5,000 tokens.\n\nHow agents ask for it: remember that Alice owns the billing service; what do we know about project X; note this decision for next time\n\n## Does the job, with a limit\n\nIt does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.\n\n- **Memory (MCP reference server) (Model Context Protocol, a Series of LF Projects, LLC)**: MEMORY_FILE_PATH sets the file, saves are atomic, a lock file serialises writers across processes (fixing #4797), and search_nodes takes limit and observationLimit. The default path still sits in node_modules, which a reinstall can delete.\n\n- not this job: Mirage (AgentVFS) (Strukto.AI): No entity or relation store; files only.\n- not this job: ClickHouse MCP Server (ClickHouse, Inc.): No entity or memory store.\n- not this job: CoinMarketCap x402 API (CoinMarketCap): No store the agent writes to.\n- not this job: DBHub (Bytebase): No entity or memory store; it queries existing databases.\n- not this job: Filesystem (MCP reference server) (Model Context Protocol, a Series of LF Projects, LLC): Reads and writes files; no entity, relation or search-by-name store.\n- not this job: MCP Toolbox for Databases (Google LLC): No memory store.\n- not this job: MongoDB MCP Server (MongoDB, Inc.): No memory or graph store of its own.\n- not this job: Nansen x402 API (Nansen): No store the agent writes to.\n- not this job: Neon MCP Server (Neon, LLC (subsidiary of Databricks, Inc.)): No memory store.\n- not this job: ObsessionDB (Kellerkind UG (haftungsbeschränkt)): No memory store.\n- not this job: PlanetScale MCP Server (PlanetScale, Inc.): No memory store.\n- not this job: Postgres MCP Pro (Crystal Corp.): No memory store.\n- not this job: Redis MCP (Redis Ltd.): Stores keys and JSON in Redis, not entities and relations in a file at a chosen path.\n- not this job: ReqRes MCP server (ReqRes): Generic JSON records, not a local graph file.\n- not this job: Snowflake-managed MCP server (Snowflake Inc.): No memory store.\n- not this job: Supabase API + MCP (Supabase Pte. Ltd.): No memory graph store.\n\n## What stops agents finishing\n\n- default file location inside a package cache that a reinstall removes\n- two processes overwrite each other's writes\n\n## Around it\n\nBefore: agent-memory.store-and-confirm. After: agent-memory.user-facts-recall.\n\n## Where this comes from\n\nWeight 7 of 100 in Databases \u0026 files: 4 of 52 accounts on the memory reference server; agent-memory has its own map, so kept small here\n\nAs data: https://www.anchorterminal.com/api/v1/use-cases/data.memory-graph.json, or the MCP tool get_use_case with id data.memory-graph. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-16",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.8",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "For agents",
        "url": "https://www.anchorterminal.com/agents/"
      },
      {
        "name": "Niches",
        "url": "https://www.anchorterminal.com/niches/"
      },
      {
        "name": "Keep facts in a small local graph the agent reads back",
        "url": ""
      }
    ],
    "description": "Keep facts in a small local graph the agent reads back: 0 companies fully do it, 1 with a limit, 17 carry the tag. Every listing in Databases \u0026 files with the reason and a source.",
    "facts": null,
    "h1": "Keep facts in a small local graph the agent reads back",
    "image": "https://www.anchorterminal.com/assets/og/default.png",
    "path": "/niches/data.memory-graph/",
    "published": "",
    "section": "agents",
    "title": "Keep facts in a small local graph the agent reads back | Anchor…",
    "toc": null,
    "updated": "2026-11-15",
    "url": "https://www.anchorterminal.com/niches/data.memory-graph/"
  },
  "tokens": {
    "markdown": 1000,
    "slim": 230
  },
  "version": 1
}
