{"slug": "we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it", "title": "We tested Memorify's MCP gateway — 12ms sync holds until it", "summary": "Memorify's MCP gateway delivers a verified 12ms sync for agent tool registration, but multi-tenant memory isolation, vector search latency beyond 100k vectors, and rigid verb protocols limit its use for complex workloads, according to a hands-on test by an unnamed engineering team. The gateway replaces ~300 lines of glue code per agent for simple use cases, but the team reports cold-start recall latency of 300-500ms and a single point of failure for OAuth connectors.", "body_md": "# We tested Memorify's MCP gateway — 12ms sync holds until it\n\n[MCP](/en/tags/mcp/)server chain we've been duct-taping together since January. The pitch is clean: one gateway, one token, every agent gets the same toolset without rebooting. The ~12ms sync claim is real — we verified it against a local Neon instance with ElectricSQL replication. But the demo stops exactly where our problems start.\n\nThe handshake flow works as advertised. Register an agent, get a scoped Bearer token, hit `/mcp`\n\nwith `tools/list`\n\nand you're live:\n\n```\ncurl -sS -X POST https://memorify.dev/mcp \\\n -H \"Content-Type: application/json\" \\\n -H \"Accept: application/json\" \\\n -H \"Authorization: Bearer $MEMORIFY_AGENT_TOKEN\" \\\n -d '{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/list\",\"params\":{}}'\n```\n\nReturns the registered toolset instantly. `whoami`\n\nconfirms identity and scopes. No argument there — onboarding friction is genuinely lower than our current `claude_desktop_config.json`\n\n+ per-IDE extension mess.\n\nHot-plugging MCP servers through the dashboard also works. Add a new server, agents see it on the next `tools/list`\n\ncall without restart. That alone would save us ~20 minutes per developer per week in context-switching overhead.\n\nWhere it gets murky:\n\n**Memory isolation.** The `/memory`\n\nverb namespace (remember, recall, update) assumes a single workspace per token. Our org runs multi-tenant workloads — same agent identity, different customer contexts. Memorify's scoping model doesn't natively partition memory by tenant without issuing separate tokens per context, which defeats the \"one connection\" promise.\n\n**Vector search at scale.** `/documents`\n\nuses pgvector under the hood. Our corpus is ~400k chunks across 12 repos. Cold-start recall latency jumped to 300-500ms once the index exceeded 100k vectors. The 12ms figure only holds for the MCP bus itself, not the semantic layer. We'd need a dedicated vector tier anyway.\n\n**OAuth connector maintenance.** The gateway abstracts OAuth flows for GitHub, Linear, Notion, etc. Convenient until a provider rotates scopes or deprecates an endpoint. Now the gateway becomes a single point of failure for *all* agents simultaneously. Our current fragmented approach at least contains blast radius.\n\n**Verb protocol rigidity.** The semantic verbs (`/skills`\n\n, `/mcp`\n\n, `/documents`\n\n) are opinionated. We have custom tooling that doesn't map cleanly — think internal deploy CLI, feature flag service, cost attribution API. Extending the verb set requires gateway changes, not just agent-side config.\n\nWe're still running a pilot with two teams on low-risk workloads (documentation ingestion, PR summarization). For those, Memorify replaces ~300 lines of glue code per agent. But for our core autonomous workflows — the ones that actually touch production — the abstraction leaks too much.\n\nIf you're running a handful of agents against a single workspace with standard SaaS integrations, it's probably worth the trade-off. If you're doing multi-tenant, high-volume, or custom tooling, budget time to hit the walls we found.\n\n[Next Testing Three Agent Frameworks on A2A — What Broke and What →](/en/threads/7071/)", "url": "https://wpnews.pro/news/we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it", "canonical_source": "https://promptcube3.com/en/threads/7093/", "published_at": "2026-08-20 20:47:35+00:00", "updated_at": "2026-08-20 21:14:45.915916+00:00", "lang": "en", "topics": ["developer-tools", "ai-agents", "ai-infrastructure"], "entities": ["Memorify", "Neon", "ElectricSQL", "GitHub", "Linear", "Notion"], "alternates": {"html": "https://wpnews.pro/news/we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it", "markdown": "https://wpnews.pro/news/we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it.md", "text": "https://wpnews.pro/news/we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it.txt", "jsonld": "https://wpnews.pro/news/we-tested-memorify-s-mcp-gateway-12ms-sync-holds-until-it.jsonld"}}