v0.1.21 — feat: add MCP Server Generator for generated apps
Release v0.1.21: MCP Server Generator for Generated Apps
February 12, 2026
What changed
- MCP Server Generator — Every app built by the platform now auto-generates a Model Context Protocol server, making it controllable by any AI agent.
- Core generator library (
src/lib/mcp/) — Regex-based Drizzle schema parser extracts table definitions, columns, enums, and references. Templates produce CRUD tools (5 per table), auth tools (2), and health tools (3). - Inngest agent (
pipeline/mcp-generate) — Deterministic code generation (no AI calls). Reads the Drizzle schema from GitHub, generates all files, creates a feature branch and PR. - tRPC API —
mcp.previewreturns a preview of tools that would be generated.mcp.generatetriggers generation. Both include auth and audit logging. - Generated server structure — Complete
mcp-server/directory with stdio entry point (Claude Desktop), HTTP/SSE entry point (remote agents), DB connection, adapted schema, and per-table tool files.
Why it matters
Generated apps are now AI-agent-ready out of the box. Any MCP-compatible client — Claude Desktop, remote agent orchestrators, or custom tooling — can interact with the app's data through auto-generated CRUD, auth, and health tools without writing integration code.
Technical notes
- New Inngest event:
pipeline/mcp-generate - New tRPC router:
mcp(added to root router) - 1,842 lines added across 10 files
- No new database tables; no breaking changes
Release Notes
Release Summary
Version: v0.1.21
Release URL: https://github.com/dwdavies/AnyApp/releases/tag/v0.1.21
PR #60 Status
The PR's changes were already integrated into the main branch, so a squash merge was not possible (GitHub reported the PR as not mergeable). However, all 10 files and 1,842 lines of additions from the PR are confirmed present on main.
What Was Released
MCP Server Generator for Generated Apps — Every app built by the platform now auto-generates a Model Context Protocol (MCP) server, making generated apps fully controllable by any AI agent.
Key components:
- Core generator library (
src/lib/mcp/) — Schema parser, code templates, and orchestrator - Inngest agent (
pipeline/mcp-generate) — Deterministic agent that reads Drizzle schemas and generates complete MCP server packages - tRPC API —
mcp.previewandmcp.generateendpoints with auth and audit logging - Generated output — Complete
mcp-server/directory with stdio + HTTP/SSE servers, CRUD tools (5 per table), auth tools, and health tools
This closes the mission gap: "No MCP Server Generator for Generated Apps".