Skip to content

What gets deployed

Resources created by the Alchemy stack for remote memory and MCP.

Updated View as Markdown
For humans

The Alchemy stack uses a router plus private workers:

Resource Role
Router Sole public entry (custom domain or workers.dev). Dispatches paths
API worker Authenticated REST (/query, /migrate, /vectorize/*, /product/*). Sole owner of the D1, Vectorize, and Workers AI bindings
MCP worker API-only gateway: MCP tools at /mcp + OAuth (/authorize, /callback, …) forwarded to /product/* over a service binding
Docs worker Optional static docs site (catch-all on the router)
D1 database Canonical memory rows and FTS
Vectorize index Semantic embeddings
Workers AI BGE embeddings for writes and semantic search
KV (OAuth) Authorization state and tokens when MCP is enabled

Public base URL is always the router. CLI and REST use /query on that host; MCP clients use /mcp on the same host after Enable MCP. Documentation pages live under /docs/* on the docs host (not under /mcp, which is reserved for the protocol).

Source of truth: iac/ (alchemy.run.ts, src/deploy-config.ts, src/router/).

Next: Deployment options.

Navigation

Type to search…

↑↓ navigate↵ selectEsc close